]> git.ipfire.org Git - thirdparty/mdadm.git/blame - super-ddf.c
Typo in dprintf
[thirdparty/mdadm.git] / super-ddf.c
CommitLineData
a322f70c
DW
1/*
2 * mdadm - manage Linux "md" devices aka RAID arrays.
3 *
e736b623 4 * Copyright (C) 2006-2009 Neil Brown <neilb@suse.de>
a322f70c
DW
5 *
6 *
7 * This program is free software; you can redistribute it and/or modify
8 * it under the terms of the GNU General Public License as published by
9 * the Free Software Foundation; either version 2 of the License, or
10 * (at your option) any later version.
11 *
12 * This program is distributed in the hope that it will be useful,
13 * but WITHOUT ANY WARRANTY; without even the implied warranty of
14 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
15 * GNU General Public License for more details.
16 *
17 * You should have received a copy of the GNU General Public License
18 * along with this program; if not, write to the Free Software
19 * Foundation, Inc., 59 Temple Place, Suite 330, Boston, MA 02111-1307 USA
20 *
21 * Author: Neil Brown
22 * Email: <neil@brown.name>
23 *
24 * Specifications for DDF takes from Common RAID DDF Specification Revision 1.2
25 * (July 28 2006). Reused by permission of SNIA.
26 */
27
28#define HAVE_STDINT_H 1
29#include "mdadm.h"
549e9569 30#include "mdmon.h"
a322f70c
DW
31#include "sha1.h"
32#include <values.h>
33
a322f70c
DW
34/* a non-official T10 name for creation GUIDs */
35static char T10[] = "Linux-MD";
36
37/* DDF timestamps are 1980 based, so we need to add
38 * second-in-decade-of-seventies to convert to linux timestamps.
39 * 10 years with 2 leap years.
40 */
41#define DECADE (3600*24*(365*10+2))
42unsigned long crc32(
43 unsigned long crc,
44 const unsigned char *buf,
45 unsigned len);
46
47/* The DDF metadata handling.
48 * DDF metadata lives at the end of the device.
49 * The last 512 byte block provides an 'anchor' which is used to locate
50 * the rest of the metadata which usually lives immediately behind the anchor.
51 *
52 * Note:
53 * - all multibyte numeric fields are bigendian.
54 * - all strings are space padded.
55 *
56 */
57
58/* Primary Raid Level (PRL) */
59#define DDF_RAID0 0x00
60#define DDF_RAID1 0x01
61#define DDF_RAID3 0x03
62#define DDF_RAID4 0x04
63#define DDF_RAID5 0x05
64#define DDF_RAID1E 0x11
65#define DDF_JBOD 0x0f
66#define DDF_CONCAT 0x1f
67#define DDF_RAID5E 0x15
68#define DDF_RAID5EE 0x25
59e36268 69#define DDF_RAID6 0x06
a322f70c
DW
70
71/* Raid Level Qualifier (RLQ) */
72#define DDF_RAID0_SIMPLE 0x00
73#define DDF_RAID1_SIMPLE 0x00 /* just 2 devices in this plex */
74#define DDF_RAID1_MULTI 0x01 /* exactly 3 devices in this plex */
75#define DDF_RAID3_0 0x00 /* parity in first extent */
76#define DDF_RAID3_N 0x01 /* parity in last extent */
77#define DDF_RAID4_0 0x00 /* parity in first extent */
78#define DDF_RAID4_N 0x01 /* parity in last extent */
79/* these apply to raid5e and raid5ee as well */
80#define DDF_RAID5_0_RESTART 0x00 /* same as 'right asymmetric' - layout 1 */
59e36268 81#define DDF_RAID6_0_RESTART 0x01 /* raid6 different from raid5 here!!! */
a322f70c
DW
82#define DDF_RAID5_N_RESTART 0x02 /* same as 'left asymmetric' - layout 0 */
83#define DDF_RAID5_N_CONTINUE 0x03 /* same as 'left symmetric' - layout 2 */
84
85#define DDF_RAID1E_ADJACENT 0x00 /* raid10 nearcopies==2 */
86#define DDF_RAID1E_OFFSET 0x01 /* raid10 offsetcopies==2 */
87
88/* Secondary RAID Level (SRL) */
89#define DDF_2STRIPED 0x00 /* This is weirder than RAID0 !! */
90#define DDF_2MIRRORED 0x01
91#define DDF_2CONCAT 0x02
92#define DDF_2SPANNED 0x03 /* This is also weird - be careful */
93
94/* Magic numbers */
95#define DDF_HEADER_MAGIC __cpu_to_be32(0xDE11DE11)
96#define DDF_CONTROLLER_MAGIC __cpu_to_be32(0xAD111111)
97#define DDF_PHYS_RECORDS_MAGIC __cpu_to_be32(0x22222222)
98#define DDF_PHYS_DATA_MAGIC __cpu_to_be32(0x33333333)
99#define DDF_VIRT_RECORDS_MAGIC __cpu_to_be32(0xDDDDDDDD)
100#define DDF_VD_CONF_MAGIC __cpu_to_be32(0xEEEEEEEE)
101#define DDF_SPARE_ASSIGN_MAGIC __cpu_to_be32(0x55555555)
102#define DDF_VU_CONF_MAGIC __cpu_to_be32(0x88888888)
103#define DDF_VENDOR_LOG_MAGIC __cpu_to_be32(0x01dBEEF0)
104#define DDF_BBM_LOG_MAGIC __cpu_to_be32(0xABADB10C)
105
106#define DDF_GUID_LEN 24
59e36268
NB
107#define DDF_REVISION_0 "01.00.00"
108#define DDF_REVISION_2 "01.02.00"
a322f70c
DW
109
110struct ddf_header {
88c164f4 111 __u32 magic; /* DDF_HEADER_MAGIC */
a322f70c
DW
112 __u32 crc;
113 char guid[DDF_GUID_LEN];
59e36268 114 char revision[8]; /* 01.02.00 */
a322f70c
DW
115 __u32 seq; /* starts at '1' */
116 __u32 timestamp;
117 __u8 openflag;
118 __u8 foreignflag;
119 __u8 enforcegroups;
120 __u8 pad0; /* 0xff */
121 __u8 pad1[12]; /* 12 * 0xff */
122 /* 64 bytes so far */
123 __u8 header_ext[32]; /* reserved: fill with 0xff */
124 __u64 primary_lba;
125 __u64 secondary_lba;
126 __u8 type;
127 __u8 pad2[3]; /* 0xff */
128 __u32 workspace_len; /* sectors for vendor space -
129 * at least 32768(sectors) */
130 __u64 workspace_lba;
131 __u16 max_pd_entries; /* one of 15, 63, 255, 1023, 4095 */
132 __u16 max_vd_entries; /* 2^(4,6,8,10,12)-1 : i.e. as above */
133 __u16 max_partitions; /* i.e. max num of configuration
134 record entries per disk */
135 __u16 config_record_len; /* 1 +ROUNDUP(max_primary_element_entries
136 *12/512) */
137 __u16 max_primary_element_entries; /* 16, 64, 256, 1024, or 4096 */
138 __u8 pad3[54]; /* 0xff */
139 /* 192 bytes so far */
140 __u32 controller_section_offset;
141 __u32 controller_section_length;
142 __u32 phys_section_offset;
143 __u32 phys_section_length;
144 __u32 virt_section_offset;
145 __u32 virt_section_length;
146 __u32 config_section_offset;
147 __u32 config_section_length;
148 __u32 data_section_offset;
149 __u32 data_section_length;
150 __u32 bbm_section_offset;
151 __u32 bbm_section_length;
152 __u32 diag_space_offset;
153 __u32 diag_space_length;
154 __u32 vendor_offset;
155 __u32 vendor_length;
156 /* 256 bytes so far */
157 __u8 pad4[256]; /* 0xff */
158};
159
160/* type field */
161#define DDF_HEADER_ANCHOR 0x00
162#define DDF_HEADER_PRIMARY 0x01
163#define DDF_HEADER_SECONDARY 0x02
164
165/* The content of the 'controller section' - global scope */
166struct ddf_controller_data {
88c164f4 167 __u32 magic; /* DDF_CONTROLLER_MAGIC */
a322f70c
DW
168 __u32 crc;
169 char guid[DDF_GUID_LEN];
170 struct controller_type {
171 __u16 vendor_id;
172 __u16 device_id;
173 __u16 sub_vendor_id;
174 __u16 sub_device_id;
175 } type;
176 char product_id[16];
177 __u8 pad[8]; /* 0xff */
178 __u8 vendor_data[448];
179};
180
181/* The content of phys_section - global scope */
182struct phys_disk {
88c164f4 183 __u32 magic; /* DDF_PHYS_RECORDS_MAGIC */
a322f70c
DW
184 __u32 crc;
185 __u16 used_pdes;
186 __u16 max_pdes;
187 __u8 pad[52];
188 struct phys_disk_entry {
189 char guid[DDF_GUID_LEN];
190 __u32 refnum;
191 __u16 type;
192 __u16 state;
193 __u64 config_size; /* DDF structures must be after here */
194 char path[18]; /* another horrible structure really */
195 __u8 pad[6];
196 } entries[0];
197};
198
199/* phys_disk_entry.type is a bitmap - bigendian remember */
200#define DDF_Forced_PD_GUID 1
201#define DDF_Active_in_VD 2
88c164f4 202#define DDF_Global_Spare 4 /* VD_CONF records are ignored */
a322f70c
DW
203#define DDF_Spare 8 /* overrides Global_spare */
204#define DDF_Foreign 16
205#define DDF_Legacy 32 /* no DDF on this device */
206
207#define DDF_Interface_mask 0xf00
208#define DDF_Interface_SCSI 0x100
209#define DDF_Interface_SAS 0x200
210#define DDF_Interface_SATA 0x300
211#define DDF_Interface_FC 0x400
212
213/* phys_disk_entry.state is a bigendian bitmap */
214#define DDF_Online 1
215#define DDF_Failed 2 /* overrides 1,4,8 */
216#define DDF_Rebuilding 4
217#define DDF_Transition 8
218#define DDF_SMART 16
219#define DDF_ReadErrors 32
220#define DDF_Missing 64
221
222/* The content of the virt_section global scope */
223struct virtual_disk {
88c164f4 224 __u32 magic; /* DDF_VIRT_RECORDS_MAGIC */
a322f70c
DW
225 __u32 crc;
226 __u16 populated_vdes;
227 __u16 max_vdes;
228 __u8 pad[52];
229 struct virtual_entry {
230 char guid[DDF_GUID_LEN];
231 __u16 unit;
232 __u16 pad0; /* 0xffff */
233 __u16 guid_crc;
234 __u16 type;
235 __u8 state;
236 __u8 init_state;
237 __u8 pad1[14];
238 char name[16];
239 } entries[0];
240};
241
242/* virtual_entry.type is a bitmap - bigendian */
243#define DDF_Shared 1
244#define DDF_Enforce_Groups 2
245#define DDF_Unicode 4
246#define DDF_Owner_Valid 8
247
248/* virtual_entry.state is a bigendian bitmap */
249#define DDF_state_mask 0x7
250#define DDF_state_optimal 0x0
251#define DDF_state_degraded 0x1
252#define DDF_state_deleted 0x2
253#define DDF_state_missing 0x3
254#define DDF_state_failed 0x4
7a7cc504 255#define DDF_state_part_optimal 0x5
a322f70c
DW
256
257#define DDF_state_morphing 0x8
258#define DDF_state_inconsistent 0x10
259
260/* virtual_entry.init_state is a bigendian bitmap */
261#define DDF_initstate_mask 0x03
262#define DDF_init_not 0x00
7a7cc504
NB
263#define DDF_init_quick 0x01 /* initialisation is progress.
264 * i.e. 'state_inconsistent' */
a322f70c
DW
265#define DDF_init_full 0x02
266
267#define DDF_access_mask 0xc0
268#define DDF_access_rw 0x00
269#define DDF_access_ro 0x80
270#define DDF_access_blocked 0xc0
271
272/* The content of the config_section - local scope
273 * It has multiple records each config_record_len sectors
274 * They can be vd_config or spare_assign
275 */
276
277struct vd_config {
88c164f4 278 __u32 magic; /* DDF_VD_CONF_MAGIC */
a322f70c
DW
279 __u32 crc;
280 char guid[DDF_GUID_LEN];
281 __u32 timestamp;
282 __u32 seqnum;
283 __u8 pad0[24];
284 __u16 prim_elmnt_count;
285 __u8 chunk_shift; /* 0 == 512, 1==1024 etc */
286 __u8 prl;
287 __u8 rlq;
288 __u8 sec_elmnt_count;
289 __u8 sec_elmnt_seq;
290 __u8 srl;
598f0d58
NB
291 __u64 blocks; /* blocks per component could be different
292 * on different component devices...(only
293 * for concat I hope) */
294 __u64 array_blocks; /* blocks in array */
a322f70c
DW
295 __u8 pad1[8];
296 __u32 spare_refs[8];
297 __u8 cache_pol[8];
298 __u8 bg_rate;
299 __u8 pad2[3];
300 __u8 pad3[52];
301 __u8 pad4[192];
302 __u8 v0[32]; /* reserved- 0xff */
303 __u8 v1[32]; /* reserved- 0xff */
304 __u8 v2[16]; /* reserved- 0xff */
305 __u8 v3[16]; /* reserved- 0xff */
306 __u8 vendor[32];
307 __u32 phys_refnum[0]; /* refnum of each disk in sequence */
308 /*__u64 lba_offset[0]; LBA offset in each phys. Note extents in a
309 bvd are always the same size */
310};
311
312/* vd_config.cache_pol[7] is a bitmap */
313#define DDF_cache_writeback 1 /* else writethrough */
314#define DDF_cache_wadaptive 2 /* only applies if writeback */
315#define DDF_cache_readahead 4
316#define DDF_cache_radaptive 8 /* only if doing read-ahead */
317#define DDF_cache_ifnobatt 16 /* even to write cache if battery is poor */
318#define DDF_cache_wallowed 32 /* enable write caching */
319#define DDF_cache_rallowed 64 /* enable read caching */
320
321struct spare_assign {
88c164f4 322 __u32 magic; /* DDF_SPARE_ASSIGN_MAGIC */
a322f70c
DW
323 __u32 crc;
324 __u32 timestamp;
325 __u8 reserved[7];
326 __u8 type;
327 __u16 populated; /* SAEs used */
328 __u16 max; /* max SAEs */
329 __u8 pad[8];
330 struct spare_assign_entry {
331 char guid[DDF_GUID_LEN];
332 __u16 secondary_element;
333 __u8 pad[6];
334 } spare_ents[0];
335};
336/* spare_assign.type is a bitmap */
337#define DDF_spare_dedicated 0x1 /* else global */
338#define DDF_spare_revertible 0x2 /* else committable */
339#define DDF_spare_active 0x4 /* else not active */
340#define DDF_spare_affinity 0x8 /* enclosure affinity */
341
342/* The data_section contents - local scope */
343struct disk_data {
88c164f4 344 __u32 magic; /* DDF_PHYS_DATA_MAGIC */
a322f70c
DW
345 __u32 crc;
346 char guid[DDF_GUID_LEN];
347 __u32 refnum; /* crc of some magic drive data ... */
348 __u8 forced_ref; /* set when above was not result of magic */
349 __u8 forced_guid; /* set if guid was forced rather than magic */
350 __u8 vendor[32];
351 __u8 pad[442];
352};
353
354/* bbm_section content */
355struct bad_block_log {
356 __u32 magic;
357 __u32 crc;
358 __u16 entry_count;
359 __u32 spare_count;
360 __u8 pad[10];
361 __u64 first_spare;
362 struct mapped_block {
363 __u64 defective_start;
364 __u32 replacement_start;
365 __u16 remap_count;
366 __u8 pad[2];
367 } entries[0];
368};
369
370/* Struct for internally holding ddf structures */
371/* The DDF structure stored on each device is potentially
372 * quite different, as some data is global and some is local.
373 * The global data is:
374 * - ddf header
375 * - controller_data
376 * - Physical disk records
377 * - Virtual disk records
378 * The local data is:
379 * - Configuration records
380 * - Physical Disk data section
381 * ( and Bad block and vendor which I don't care about yet).
382 *
383 * The local data is parsed into separate lists as it is read
384 * and reconstructed for writing. This means that we only need
385 * to make config changes once and they are automatically
386 * propagated to all devices.
387 * Note that the ddf_super has space of the conf and disk data
388 * for this disk and also for a list of all such data.
389 * The list is only used for the superblock that is being
390 * built in Create or Assemble to describe the whole array.
391 */
392struct ddf_super {
6416d527 393 struct ddf_header anchor, primary, secondary;
a322f70c 394 struct ddf_controller_data controller;
6416d527 395 struct ddf_header *active;
a322f70c
DW
396 struct phys_disk *phys;
397 struct virtual_disk *virt;
398 int pdsize, vdsize;
f21e18ca 399 unsigned int max_part, mppe, conf_rec_len;
d2ca6449 400 int currentdev;
18a2f463 401 int updates_pending;
a322f70c 402 struct vcl {
6416d527
NB
403 union {
404 char space[512];
405 struct {
406 struct vcl *next;
407 __u64 *lba_offset; /* location in 'conf' of
408 * the lba table */
f21e18ca 409 unsigned int vcnum; /* index into ->virt */
6416d527
NB
410 __u64 *block_sizes; /* NULL if all the same */
411 };
412 };
a322f70c 413 struct vd_config conf;
d2ca6449 414 } *conflist, *currentconf;
a322f70c 415 struct dl {
6416d527
NB
416 union {
417 char space[512];
418 struct {
419 struct dl *next;
420 int major, minor;
421 char *devname;
422 int fd;
423 unsigned long long size; /* sectors */
424 int pdnum; /* index in ->phys */
425 struct spare_assign *spare;
8592f29d
N
426 void *mdupdate; /* hold metadata update */
427
428 /* These fields used by auto-layout */
429 int raiddisk; /* slot to fill in autolayout */
430 __u64 esize;
6416d527
NB
431 };
432 };
a322f70c 433 struct disk_data disk;
b2280677 434 struct vcl *vlist[0]; /* max_part in size */
2cc2983d 435 } *dlist, *add_list;
a322f70c
DW
436};
437
438#ifndef offsetof
439#define offsetof(t,f) ((size_t)&(((t*)0)->f))
440#endif
441
a322f70c 442
f21e18ca 443static unsigned int calc_crc(void *buf, int len)
a322f70c
DW
444{
445 /* crcs are always at the same place as in the ddf_header */
446 struct ddf_header *ddf = buf;
447 __u32 oldcrc = ddf->crc;
448 __u32 newcrc;
449 ddf->crc = 0xffffffff;
450
451 newcrc = crc32(0, buf, len);
452 ddf->crc = oldcrc;
4abe6b70
N
453 /* The crc is store (like everything) bigendian, so convert
454 * here for simplicity
455 */
456 return __cpu_to_be32(newcrc);
a322f70c
DW
457}
458
459static int load_ddf_header(int fd, unsigned long long lba,
460 unsigned long long size,
461 int type,
462 struct ddf_header *hdr, struct ddf_header *anchor)
463{
464 /* read a ddf header (primary or secondary) from fd/lba
465 * and check that it is consistent with anchor
466 * Need to check:
467 * magic, crc, guid, rev, and LBA's header_type, and
468 * everything after header_type must be the same
469 */
470 if (lba >= size-1)
471 return 0;
472
473 if (lseek64(fd, lba<<9, 0) < 0)
474 return 0;
475
476 if (read(fd, hdr, 512) != 512)
477 return 0;
478
479 if (hdr->magic != DDF_HEADER_MAGIC)
480 return 0;
481 if (calc_crc(hdr, 512) != hdr->crc)
482 return 0;
483 if (memcmp(anchor->guid, hdr->guid, DDF_GUID_LEN) != 0 ||
484 memcmp(anchor->revision, hdr->revision, 8) != 0 ||
485 anchor->primary_lba != hdr->primary_lba ||
486 anchor->secondary_lba != hdr->secondary_lba ||
487 hdr->type != type ||
488 memcmp(anchor->pad2, hdr->pad2, 512 -
489 offsetof(struct ddf_header, pad2)) != 0)
490 return 0;
491
492 /* Looks good enough to me... */
493 return 1;
494}
495
496static void *load_section(int fd, struct ddf_super *super, void *buf,
497 __u32 offset_be, __u32 len_be, int check)
498{
499 unsigned long long offset = __be32_to_cpu(offset_be);
500 unsigned long long len = __be32_to_cpu(len_be);
501 int dofree = (buf == NULL);
502
503 if (check)
504 if (len != 2 && len != 8 && len != 32
505 && len != 128 && len != 512)
506 return NULL;
507
508 if (len > 1024)
509 return NULL;
510 if (buf) {
511 /* All pre-allocated sections are a single block */
512 if (len != 1)
513 return NULL;
3d2c4fc7
DW
514 } else if (posix_memalign(&buf, 512, len<<9) != 0)
515 buf = NULL;
6416d527 516
a322f70c
DW
517 if (!buf)
518 return NULL;
519
520 if (super->active->type == 1)
521 offset += __be64_to_cpu(super->active->primary_lba);
522 else
523 offset += __be64_to_cpu(super->active->secondary_lba);
524
f21e18ca 525 if ((unsigned long long)lseek64(fd, offset<<9, 0) != (offset<<9)) {
a322f70c
DW
526 if (dofree)
527 free(buf);
528 return NULL;
529 }
f21e18ca 530 if ((unsigned long long)read(fd, buf, len<<9) != (len<<9)) {
a322f70c
DW
531 if (dofree)
532 free(buf);
533 return NULL;
534 }
535 return buf;
536}
537
538static int load_ddf_headers(int fd, struct ddf_super *super, char *devname)
539{
540 unsigned long long dsize;
541
542 get_dev_size(fd, NULL, &dsize);
543
544 if (lseek64(fd, dsize-512, 0) < 0) {
545 if (devname)
546 fprintf(stderr,
547 Name": Cannot seek to anchor block on %s: %s\n",
548 devname, strerror(errno));
549 return 1;
550 }
551 if (read(fd, &super->anchor, 512) != 512) {
552 if (devname)
553 fprintf(stderr,
554 Name ": Cannot read anchor block on %s: %s\n",
555 devname, strerror(errno));
556 return 1;
557 }
558 if (super->anchor.magic != DDF_HEADER_MAGIC) {
559 if (devname)
560 fprintf(stderr, Name ": no DDF anchor found on %s\n",
561 devname);
562 return 2;
563 }
564 if (calc_crc(&super->anchor, 512) != super->anchor.crc) {
565 if (devname)
566 fprintf(stderr, Name ": bad CRC on anchor on %s\n",
567 devname);
568 return 2;
569 }
59e36268
NB
570 if (memcmp(super->anchor.revision, DDF_REVISION_0, 8) != 0 &&
571 memcmp(super->anchor.revision, DDF_REVISION_2, 8) != 0) {
a322f70c
DW
572 if (devname)
573 fprintf(stderr, Name ": can only support super revision"
59e36268
NB
574 " %.8s and earlier, not %.8s on %s\n",
575 DDF_REVISION_2, super->anchor.revision,devname);
a322f70c
DW
576 return 2;
577 }
578 if (load_ddf_header(fd, __be64_to_cpu(super->anchor.primary_lba),
579 dsize >> 9, 1,
580 &super->primary, &super->anchor) == 0) {
581 if (devname)
582 fprintf(stderr,
583 Name ": Failed to load primary DDF header "
584 "on %s\n", devname);
585 return 2;
586 }
587 super->active = &super->primary;
588 if (load_ddf_header(fd, __be64_to_cpu(super->anchor.secondary_lba),
589 dsize >> 9, 2,
590 &super->secondary, &super->anchor)) {
591 if ((__be32_to_cpu(super->primary.seq)
592 < __be32_to_cpu(super->secondary.seq) &&
593 !super->secondary.openflag)
594 || (__be32_to_cpu(super->primary.seq)
595 == __be32_to_cpu(super->secondary.seq) &&
596 super->primary.openflag && !super->secondary.openflag)
597 )
598 super->active = &super->secondary;
599 }
600 return 0;
601}
602
603static int load_ddf_global(int fd, struct ddf_super *super, char *devname)
604{
605 void *ok;
606 ok = load_section(fd, super, &super->controller,
607 super->active->controller_section_offset,
608 super->active->controller_section_length,
609 0);
610 super->phys = load_section(fd, super, NULL,
611 super->active->phys_section_offset,
612 super->active->phys_section_length,
613 1);
614 super->pdsize = __be32_to_cpu(super->active->phys_section_length) * 512;
615
616 super->virt = load_section(fd, super, NULL,
617 super->active->virt_section_offset,
618 super->active->virt_section_length,
619 1);
620 super->vdsize = __be32_to_cpu(super->active->virt_section_length) * 512;
621 if (!ok ||
622 !super->phys ||
623 !super->virt) {
624 free(super->phys);
625 free(super->virt);
a2349791
NB
626 super->phys = NULL;
627 super->virt = NULL;
a322f70c
DW
628 return 2;
629 }
630 super->conflist = NULL;
631 super->dlist = NULL;
8c3b8c2c
NB
632
633 super->max_part = __be16_to_cpu(super->active->max_partitions);
634 super->mppe = __be16_to_cpu(super->active->max_primary_element_entries);
635 super->conf_rec_len = __be16_to_cpu(super->active->config_record_len);
a322f70c
DW
636 return 0;
637}
638
639static int load_ddf_local(int fd, struct ddf_super *super,
640 char *devname, int keep)
641{
642 struct dl *dl;
643 struct stat stb;
644 char *conf;
f21e18ca
N
645 unsigned int i;
646 unsigned int confsec;
b2280677 647 int vnum;
f21e18ca 648 unsigned int max_virt_disks = __be16_to_cpu(super->active->max_vd_entries);
d2ca6449 649 unsigned long long dsize;
a322f70c
DW
650
651 /* First the local disk info */
3d2c4fc7 652 if (posix_memalign((void**)&dl, 512,
6416d527 653 sizeof(*dl) +
3d2c4fc7
DW
654 (super->max_part) * sizeof(dl->vlist[0])) != 0) {
655 fprintf(stderr, Name ": %s could not allocate disk info buffer\n",
656 __func__);
657 return 1;
658 }
a322f70c
DW
659
660 load_section(fd, super, &dl->disk,
661 super->active->data_section_offset,
662 super->active->data_section_length,
663 0);
664 dl->devname = devname ? strdup(devname) : NULL;
598f0d58 665
a322f70c
DW
666 fstat(fd, &stb);
667 dl->major = major(stb.st_rdev);
668 dl->minor = minor(stb.st_rdev);
669 dl->next = super->dlist;
670 dl->fd = keep ? fd : -1;
d2ca6449
NB
671
672 dl->size = 0;
673 if (get_dev_size(fd, devname, &dsize))
674 dl->size = dsize >> 9;
b2280677 675 dl->spare = NULL;
f21e18ca 676 for (i = 0 ; i < super->max_part ; i++)
a322f70c
DW
677 dl->vlist[i] = NULL;
678 super->dlist = dl;
59e36268 679 dl->pdnum = -1;
f21e18ca 680 for (i = 0; i < __be16_to_cpu(super->active->max_pd_entries); i++)
5575e7d9
NB
681 if (memcmp(super->phys->entries[i].guid,
682 dl->disk.guid, DDF_GUID_LEN) == 0)
683 dl->pdnum = i;
684
a322f70c
DW
685 /* Now the config list. */
686 /* 'conf' is an array of config entries, some of which are
687 * probably invalid. Those which are good need to be copied into
688 * the conflist
689 */
a322f70c
DW
690
691 conf = load_section(fd, super, NULL,
692 super->active->config_section_offset,
693 super->active->config_section_length,
694 0);
695
b2280677 696 vnum = 0;
e223334f
N
697 for (confsec = 0;
698 confsec < __be32_to_cpu(super->active->config_section_length);
699 confsec += super->conf_rec_len) {
a322f70c 700 struct vd_config *vd =
e223334f 701 (struct vd_config *)((char*)conf + confsec*512);
a322f70c
DW
702 struct vcl *vcl;
703
b2280677
NB
704 if (vd->magic == DDF_SPARE_ASSIGN_MAGIC) {
705 if (dl->spare)
706 continue;
3d2c4fc7
DW
707 if (posix_memalign((void**)&dl->spare, 512,
708 super->conf_rec_len*512) != 0) {
709 fprintf(stderr, Name
710 ": %s could not allocate spare info buf\n",
711 __func__);
712 return 1;
713 }
714
b2280677
NB
715 memcpy(dl->spare, vd, super->conf_rec_len*512);
716 continue;
717 }
a322f70c
DW
718 if (vd->magic != DDF_VD_CONF_MAGIC)
719 continue;
720 for (vcl = super->conflist; vcl; vcl = vcl->next) {
721 if (memcmp(vcl->conf.guid,
722 vd->guid, DDF_GUID_LEN) == 0)
723 break;
724 }
725
726 if (vcl) {
b2280677 727 dl->vlist[vnum++] = vcl;
a322f70c
DW
728 if (__be32_to_cpu(vd->seqnum) <=
729 __be32_to_cpu(vcl->conf.seqnum))
730 continue;
59e36268 731 } else {
3d2c4fc7 732 if (posix_memalign((void**)&vcl, 512,
6416d527 733 (super->conf_rec_len*512 +
3d2c4fc7
DW
734 offsetof(struct vcl, conf))) != 0) {
735 fprintf(stderr, Name
736 ": %s could not allocate vcl buf\n",
737 __func__);
738 return 1;
739 }
a322f70c 740 vcl->next = super->conflist;
59e36268 741 vcl->block_sizes = NULL; /* FIXME not for CONCAT */
a322f70c 742 super->conflist = vcl;
b2280677 743 dl->vlist[vnum++] = vcl;
a322f70c 744 }
8c3b8c2c 745 memcpy(&vcl->conf, vd, super->conf_rec_len*512);
a322f70c 746 vcl->lba_offset = (__u64*)
8c3b8c2c 747 &vcl->conf.phys_refnum[super->mppe];
59e36268
NB
748
749 for (i=0; i < max_virt_disks ; i++)
750 if (memcmp(super->virt->entries[i].guid,
751 vcl->conf.guid, DDF_GUID_LEN)==0)
752 break;
753 if (i < max_virt_disks)
754 vcl->vcnum = i;
a322f70c
DW
755 }
756 free(conf);
757
758 return 0;
759}
760
761#ifndef MDASSEMBLE
762static int load_super_ddf_all(struct supertype *st, int fd,
e1902a7b 763 void **sbp, char *devname);
a322f70c 764#endif
37424f13
DW
765
766static void free_super_ddf(struct supertype *st);
767
a322f70c
DW
768static int load_super_ddf(struct supertype *st, int fd,
769 char *devname)
770{
771 unsigned long long dsize;
772 struct ddf_super *super;
773 int rv;
774
a322f70c
DW
775 if (get_dev_size(fd, devname, &dsize) == 0)
776 return 1;
777
691c6ee1
N
778 if (test_partition(fd))
779 /* DDF is not allowed on partitions */
780 return 1;
781
a322f70c
DW
782 /* 32M is a lower bound */
783 if (dsize <= 32*1024*1024) {
97320d7c 784 if (devname)
a322f70c
DW
785 fprintf(stderr,
786 Name ": %s is too small for ddf: "
787 "size is %llu sectors.\n",
788 devname, dsize>>9);
97320d7c 789 return 1;
a322f70c
DW
790 }
791 if (dsize & 511) {
97320d7c 792 if (devname)
a322f70c
DW
793 fprintf(stderr,
794 Name ": %s is an odd size for ddf: "
795 "size is %llu bytes.\n",
796 devname, dsize);
97320d7c 797 return 1;
a322f70c
DW
798 }
799
37424f13
DW
800 free_super_ddf(st);
801
6416d527 802 if (posix_memalign((void**)&super, 512, sizeof(*super))!= 0) {
a322f70c
DW
803 fprintf(stderr, Name ": malloc of %zu failed.\n",
804 sizeof(*super));
805 return 1;
806 }
a2349791 807 memset(super, 0, sizeof(*super));
a322f70c
DW
808
809 rv = load_ddf_headers(fd, super, devname);
810 if (rv) {
811 free(super);
812 return rv;
813 }
814
815 /* Have valid headers and have chosen the best. Let's read in the rest*/
816
817 rv = load_ddf_global(fd, super, devname);
818
819 if (rv) {
820 if (devname)
821 fprintf(stderr,
822 Name ": Failed to load all information "
823 "sections on %s\n", devname);
824 free(super);
825 return rv;
826 }
827
3d2c4fc7
DW
828 rv = load_ddf_local(fd, super, devname, 0);
829
830 if (rv) {
831 if (devname)
832 fprintf(stderr,
833 Name ": Failed to load all information "
834 "sections on %s\n", devname);
835 free(super);
836 return rv;
837 }
a322f70c
DW
838
839 /* Should possibly check the sections .... */
840
841 st->sb = super;
842 if (st->ss == NULL) {
843 st->ss = &super_ddf;
844 st->minor_version = 0;
845 st->max_devs = 512;
846 }
847 return 0;
848
849}
850
851static void free_super_ddf(struct supertype *st)
852{
853 struct ddf_super *ddf = st->sb;
854 if (ddf == NULL)
855 return;
856 free(ddf->phys);
857 free(ddf->virt);
858 while (ddf->conflist) {
859 struct vcl *v = ddf->conflist;
860 ddf->conflist = v->next;
59e36268
NB
861 if (v->block_sizes)
862 free(v->block_sizes);
a322f70c
DW
863 free(v);
864 }
865 while (ddf->dlist) {
866 struct dl *d = ddf->dlist;
867 ddf->dlist = d->next;
868 if (d->fd >= 0)
869 close(d->fd);
b2280677
NB
870 if (d->spare)
871 free(d->spare);
a322f70c
DW
872 free(d);
873 }
874 free(ddf);
875 st->sb = NULL;
876}
877
878static struct supertype *match_metadata_desc_ddf(char *arg)
879{
880 /* 'ddf' only support containers */
881 struct supertype *st;
882 if (strcmp(arg, "ddf") != 0 &&
883 strcmp(arg, "default") != 0
884 )
885 return NULL;
886
887 st = malloc(sizeof(*st));
ef609477 888 memset(st, 0, sizeof(*st));
d1d599ea 889 st->container_dev = NoMdDev;
a322f70c
DW
890 st->ss = &super_ddf;
891 st->max_devs = 512;
892 st->minor_version = 0;
893 st->sb = NULL;
894 return st;
895}
896
a322f70c
DW
897
898#ifndef MDASSEMBLE
899
900static mapping_t ddf_state[] = {
901 { "Optimal", 0},
902 { "Degraded", 1},
903 { "Deleted", 2},
904 { "Missing", 3},
905 { "Failed", 4},
906 { "Partially Optimal", 5},
907 { "-reserved-", 6},
908 { "-reserved-", 7},
909 { NULL, 0}
910};
911
912static mapping_t ddf_init_state[] = {
913 { "Not Initialised", 0},
914 { "QuickInit in Progress", 1},
915 { "Fully Initialised", 2},
916 { "*UNKNOWN*", 3},
917 { NULL, 0}
918};
919static mapping_t ddf_access[] = {
920 { "Read/Write", 0},
921 { "Reserved", 1},
922 { "Read Only", 2},
923 { "Blocked (no access)", 3},
924 { NULL ,0}
925};
926
927static mapping_t ddf_level[] = {
928 { "RAID0", DDF_RAID0},
929 { "RAID1", DDF_RAID1},
930 { "RAID3", DDF_RAID3},
931 { "RAID4", DDF_RAID4},
932 { "RAID5", DDF_RAID5},
933 { "RAID1E",DDF_RAID1E},
934 { "JBOD", DDF_JBOD},
935 { "CONCAT",DDF_CONCAT},
936 { "RAID5E",DDF_RAID5E},
937 { "RAID5EE",DDF_RAID5EE},
938 { "RAID6", DDF_RAID6},
939 { NULL, 0}
940};
941static mapping_t ddf_sec_level[] = {
942 { "Striped", DDF_2STRIPED},
943 { "Mirrored", DDF_2MIRRORED},
944 { "Concat", DDF_2CONCAT},
945 { "Spanned", DDF_2SPANNED},
946 { NULL, 0}
947};
948#endif
949
950struct num_mapping {
951 int num1, num2;
952};
953static struct num_mapping ddf_level_num[] = {
954 { DDF_RAID0, 0 },
955 { DDF_RAID1, 1 },
956 { DDF_RAID3, LEVEL_UNSUPPORTED },
60f18132
NB
957 { DDF_RAID4, 4 },
958 { DDF_RAID5, 5 },
a322f70c
DW
959 { DDF_RAID1E, LEVEL_UNSUPPORTED },
960 { DDF_JBOD, LEVEL_UNSUPPORTED },
961 { DDF_CONCAT, LEVEL_LINEAR },
962 { DDF_RAID5E, LEVEL_UNSUPPORTED },
963 { DDF_RAID5EE, LEVEL_UNSUPPORTED },
964 { DDF_RAID6, 6},
965 { MAXINT, MAXINT }
966};
967
968static int map_num1(struct num_mapping *map, int num)
969{
970 int i;
971 for (i=0 ; map[i].num1 != MAXINT; i++)
972 if (map[i].num1 == num)
973 break;
974 return map[i].num2;
975}
976
42dc2744
N
977static int all_ff(char *guid)
978{
979 int i;
980 for (i = 0; i < DDF_GUID_LEN; i++)
981 if (guid[i] != (char)0xff)
982 return 0;
983 return 1;
984}
985
a322f70c
DW
986#ifndef MDASSEMBLE
987static void print_guid(char *guid, int tstamp)
988{
989 /* A GUIDs are part (or all) ASCII and part binary.
990 * They tend to be space padded.
59e36268
NB
991 * We print the GUID in HEX, then in parentheses add
992 * any initial ASCII sequence, and a possible
993 * time stamp from bytes 16-19
a322f70c
DW
994 */
995 int l = DDF_GUID_LEN;
996 int i;
59e36268
NB
997
998 for (i=0 ; i<DDF_GUID_LEN ; i++) {
999 if ((i&3)==0 && i != 0) printf(":");
1000 printf("%02X", guid[i]&255);
1001 }
1002
cfccea8c 1003 printf("\n (");
a322f70c
DW
1004 while (l && guid[l-1] == ' ')
1005 l--;
1006 for (i=0 ; i<l ; i++) {
1007 if (guid[i] >= 0x20 && guid[i] < 0x7f)
1008 fputc(guid[i], stdout);
1009 else
59e36268 1010 break;
a322f70c
DW
1011 }
1012 if (tstamp) {
1013 time_t then = __be32_to_cpu(*(__u32*)(guid+16)) + DECADE;
1014 char tbuf[100];
1015 struct tm *tm;
1016 tm = localtime(&then);
59e36268 1017 strftime(tbuf, 100, " %D %T",tm);
a322f70c
DW
1018 fputs(tbuf, stdout);
1019 }
59e36268 1020 printf(")");
a322f70c
DW
1021}
1022
1023static void examine_vd(int n, struct ddf_super *sb, char *guid)
1024{
8c3b8c2c 1025 int crl = sb->conf_rec_len;
a322f70c
DW
1026 struct vcl *vcl;
1027
1028 for (vcl = sb->conflist ; vcl ; vcl = vcl->next) {
f21e18ca 1029 unsigned int i;
a322f70c
DW
1030 struct vd_config *vc = &vcl->conf;
1031
1032 if (calc_crc(vc, crl*512) != vc->crc)
1033 continue;
1034 if (memcmp(vc->guid, guid, DDF_GUID_LEN) != 0)
1035 continue;
1036
1037 /* Ok, we know about this VD, let's give more details */
b06e3095 1038 printf(" Raid Devices[%d] : %d (", n,
a322f70c 1039 __be16_to_cpu(vc->prim_elmnt_count));
f21e18ca 1040 for (i = 0; i < __be16_to_cpu(vc->prim_elmnt_count); i++) {
b06e3095
N
1041 int j;
1042 int cnt = __be16_to_cpu(sb->phys->used_pdes);
1043 for (j=0; j<cnt; j++)
1044 if (vc->phys_refnum[i] == sb->phys->entries[j].refnum)
1045 break;
1046 if (i) printf(" ");
1047 if (j < cnt)
1048 printf("%d", j);
1049 else
1050 printf("--");
1051 }
1052 printf(")\n");
1053 if (vc->chunk_shift != 255)
a322f70c
DW
1054 printf(" Chunk Size[%d] : %d sectors\n", n,
1055 1 << vc->chunk_shift);
1056 printf(" Raid Level[%d] : %s\n", n,
1057 map_num(ddf_level, vc->prl)?:"-unknown-");
1058 if (vc->sec_elmnt_count != 1) {
1059 printf(" Secondary Position[%d] : %d of %d\n", n,
1060 vc->sec_elmnt_seq, vc->sec_elmnt_count);
1061 printf(" Secondary Level[%d] : %s\n", n,
1062 map_num(ddf_sec_level, vc->srl) ?: "-unknown-");
1063 }
1064 printf(" Device Size[%d] : %llu\n", n,
c9b6907b 1065 (unsigned long long)__be64_to_cpu(vc->blocks)/2);
a322f70c 1066 printf(" Array Size[%d] : %llu\n", n,
c9b6907b 1067 (unsigned long long)__be64_to_cpu(vc->array_blocks)/2);
a322f70c
DW
1068 }
1069}
1070
1071static void examine_vds(struct ddf_super *sb)
1072{
1073 int cnt = __be16_to_cpu(sb->virt->populated_vdes);
1074 int i;
1075 printf(" Virtual Disks : %d\n", cnt);
1076
1077 for (i=0; i<cnt; i++) {
1078 struct virtual_entry *ve = &sb->virt->entries[i];
b06e3095 1079 printf("\n");
a322f70c
DW
1080 printf(" VD GUID[%d] : ", i); print_guid(ve->guid, 1);
1081 printf("\n");
1082 printf(" unit[%d] : %d\n", i, __be16_to_cpu(ve->unit));
1083 printf(" state[%d] : %s, %s%s\n", i,
1084 map_num(ddf_state, ve->state & 7),
1085 (ve->state & 8) ? "Morphing, ": "",
1086 (ve->state & 16)? "Not Consistent" : "Consistent");
1087 printf(" init state[%d] : %s\n", i,
1088 map_num(ddf_init_state, ve->init_state&3));
1089 printf(" access[%d] : %s\n", i,
1090 map_num(ddf_access, (ve->init_state>>6) & 3));
1091 printf(" Name[%d] : %.16s\n", i, ve->name);
1092 examine_vd(i, sb, ve->guid);
1093 }
1094 if (cnt) printf("\n");
1095}
1096
1097static void examine_pds(struct ddf_super *sb)
1098{
1099 int cnt = __be16_to_cpu(sb->phys->used_pdes);
1100 int i;
1101 struct dl *dl;
1102 printf(" Physical Disks : %d\n", cnt);
962371a5 1103 printf(" Number RefNo Size Device Type/State\n");
a322f70c
DW
1104
1105 for (i=0 ; i<cnt ; i++) {
1106 struct phys_disk_entry *pd = &sb->phys->entries[i];
1107 int type = __be16_to_cpu(pd->type);
1108 int state = __be16_to_cpu(pd->state);
1109
b06e3095
N
1110 //printf(" PD GUID[%d] : ", i); print_guid(pd->guid, 0);
1111 //printf("\n");
1112 printf(" %3d %08x ", i,
a322f70c 1113 __be32_to_cpu(pd->refnum));
c9b6907b
DL
1114 printf("%8lluK ",
1115 (unsigned long long)__be64_to_cpu(pd->config_size)>>1);
b06e3095
N
1116 for (dl = sb->dlist; dl ; dl = dl->next) {
1117 if (dl->disk.refnum == pd->refnum) {
1118 char *dv = map_dev(dl->major, dl->minor, 0);
1119 if (dv) {
962371a5 1120 printf("%-15s", dv);
b06e3095
N
1121 break;
1122 }
1123 }
1124 }
1125 if (!dl)
962371a5 1126 printf("%15s","");
b06e3095 1127 printf(" %s%s%s%s%s",
a322f70c 1128 (type&2) ? "active":"",
b06e3095 1129 (type&4) ? "Global-Spare":"",
a322f70c
DW
1130 (type&8) ? "spare" : "",
1131 (type&16)? ", foreign" : "",
1132 (type&32)? "pass-through" : "");
b06e3095 1133 printf("/%s%s%s%s%s%s%s",
a322f70c
DW
1134 (state&1)? "Online": "Offline",
1135 (state&2)? ", Failed": "",
1136 (state&4)? ", Rebuilding": "",
1137 (state&8)? ", in-transition": "",
b06e3095
N
1138 (state&16)? ", SMART-errors": "",
1139 (state&32)? ", Unrecovered-Read-Errors": "",
a322f70c 1140 (state&64)? ", Missing" : "");
a322f70c
DW
1141 printf("\n");
1142 }
1143}
1144
1145static void examine_super_ddf(struct supertype *st, char *homehost)
1146{
1147 struct ddf_super *sb = st->sb;
1148
1149 printf(" Magic : %08x\n", __be32_to_cpu(sb->anchor.magic));
1150 printf(" Version : %.8s\n", sb->anchor.revision);
598f0d58
NB
1151 printf("Controller GUID : "); print_guid(sb->controller.guid, 0);
1152 printf("\n");
1153 printf(" Container GUID : "); print_guid(sb->anchor.guid, 1);
a322f70c
DW
1154 printf("\n");
1155 printf(" Seq : %08x\n", __be32_to_cpu(sb->active->seq));
1156 printf(" Redundant hdr : %s\n", sb->secondary.magic == DDF_HEADER_MAGIC
1157 ?"yes" : "no");
1158 examine_vds(sb);
1159 examine_pds(sb);
1160}
1161
a5d85af7 1162static void getinfo_super_ddf(struct supertype *st, struct mdinfo *info, char *map);
ff54de6e 1163
42dc2744 1164static void uuid_from_super_ddf(struct supertype *st, int uuid[4]);
ff54de6e 1165
061f2c6a 1166static void brief_examine_super_ddf(struct supertype *st, int verbose)
4737ae25
N
1167{
1168 /* We just write a generic DDF ARRAY entry
1169 */
1170 struct mdinfo info;
1171 char nbuf[64];
a5d85af7 1172 getinfo_super_ddf(st, &info, NULL);
4737ae25
N
1173 fname_from_uuid(st, &info, nbuf, ':');
1174
1175 printf("ARRAY metadata=ddf UUID=%s\n", nbuf + 5);
1176}
1177
1178static void brief_examine_subarrays_ddf(struct supertype *st, int verbose)
a322f70c
DW
1179{
1180 /* We just write a generic DDF ARRAY entry
a322f70c 1181 */
42dc2744 1182 struct ddf_super *ddf = st->sb;
ff54de6e 1183 struct mdinfo info;
f21e18ca 1184 unsigned int i;
ff54de6e 1185 char nbuf[64];
a5d85af7 1186 getinfo_super_ddf(st, &info, NULL);
ff54de6e 1187 fname_from_uuid(st, &info, nbuf, ':');
42dc2744 1188
f21e18ca 1189 for (i = 0; i < __be16_to_cpu(ddf->virt->max_vdes); i++) {
42dc2744
N
1190 struct virtual_entry *ve = &ddf->virt->entries[i];
1191 struct vcl vcl;
1192 char nbuf1[64];
1193 if (all_ff(ve->guid))
1194 continue;
1195 memcpy(vcl.conf.guid, ve->guid, DDF_GUID_LEN);
1196 ddf->currentconf =&vcl;
1197 uuid_from_super_ddf(st, info.uuid);
1198 fname_from_uuid(st, &info, nbuf1, ':');
1199 printf("ARRAY container=%s member=%d UUID=%s\n",
1200 nbuf+5, i, nbuf1+5);
1201 }
a322f70c
DW
1202}
1203
bceedeec
N
1204static void export_examine_super_ddf(struct supertype *st)
1205{
1206 struct mdinfo info;
1207 char nbuf[64];
a5d85af7 1208 getinfo_super_ddf(st, &info, NULL);
bceedeec
N
1209 fname_from_uuid(st, &info, nbuf, ':');
1210 printf("MD_METADATA=ddf\n");
1211 printf("MD_LEVEL=container\n");
1212 printf("MD_UUID=%s\n", nbuf+5);
1213}
1214
1215
a322f70c
DW
1216static void detail_super_ddf(struct supertype *st, char *homehost)
1217{
1218 /* FIXME later
1219 * Could print DDF GUID
1220 * Need to find which array
1221 * If whole, briefly list all arrays
1222 * If one, give name
1223 */
1224}
1225
1226static void brief_detail_super_ddf(struct supertype *st)
1227{
1228 /* FIXME I really need to know which array we are detailing.
1229 * Can that be stored in ddf_super??
1230 */
1231// struct ddf_super *ddf = st->sb;
ff54de6e
N
1232 struct mdinfo info;
1233 char nbuf[64];
a5d85af7 1234 getinfo_super_ddf(st, &info, NULL);
ff54de6e
N
1235 fname_from_uuid(st, &info, nbuf,':');
1236 printf(" UUID=%s", nbuf + 5);
a322f70c 1237}
a322f70c
DW
1238#endif
1239
1240static int match_home_ddf(struct supertype *st, char *homehost)
1241{
1242 /* It matches 'this' host if the controller is a
1243 * Linux-MD controller with vendor_data matching
1244 * the hostname
1245 */
1246 struct ddf_super *ddf = st->sb;
f21e18ca 1247 unsigned int len;
d1d3482b
N
1248
1249 if (!homehost)
1250 return 0;
1251 len = strlen(homehost);
a322f70c
DW
1252
1253 return (memcmp(ddf->controller.guid, T10, 8) == 0 &&
1254 len < sizeof(ddf->controller.vendor_data) &&
1255 memcmp(ddf->controller.vendor_data, homehost,len) == 0 &&
1256 ddf->controller.vendor_data[len] == 0);
1257}
1258
0e600426 1259#ifndef MDASSEMBLE
f21e18ca 1260static struct vd_config *find_vdcr(struct ddf_super *ddf, unsigned int inst)
a322f70c 1261{
7a7cc504 1262 struct vcl *v;
59e36268 1263
7a7cc504 1264 for (v = ddf->conflist; v; v = v->next)
59e36268 1265 if (inst == v->vcnum)
7a7cc504
NB
1266 return &v->conf;
1267 return NULL;
1268}
0e600426 1269#endif
7a7cc504
NB
1270
1271static int find_phys(struct ddf_super *ddf, __u32 phys_refnum)
1272{
1273 /* Find the entry in phys_disk which has the given refnum
1274 * and return it's index
1275 */
f21e18ca
N
1276 unsigned int i;
1277 for (i = 0; i < __be16_to_cpu(ddf->phys->max_pdes); i++)
7a7cc504
NB
1278 if (ddf->phys->entries[i].refnum == phys_refnum)
1279 return i;
1280 return -1;
a322f70c
DW
1281}
1282
1283static void uuid_from_super_ddf(struct supertype *st, int uuid[4])
1284{
1285 /* The uuid returned here is used for:
1286 * uuid to put into bitmap file (Create, Grow)
1287 * uuid for backup header when saving critical section (Grow)
1288 * comparing uuids when re-adding a device into an array
51006d85
N
1289 * In these cases the uuid required is that of the data-array,
1290 * not the device-set.
1291 * uuid to recognise same set when adding a missing device back
1292 * to an array. This is a uuid for the device-set.
1293 *
a322f70c
DW
1294 * For each of these we can make do with a truncated
1295 * or hashed uuid rather than the original, as long as
1296 * everyone agrees.
a322f70c
DW
1297 * In the case of SVD we assume the BVD is of interest,
1298 * though that might be the case if a bitmap were made for
1299 * a mirrored SVD - worry about that later.
1300 * So we need to find the VD configuration record for the
1301 * relevant BVD and extract the GUID and Secondary_Element_Seq.
1302 * The first 16 bytes of the sha1 of these is used.
1303 */
1304 struct ddf_super *ddf = st->sb;
d2ca6449 1305 struct vcl *vcl = ddf->currentconf;
c5afc314
N
1306 char *guid;
1307 char buf[20];
1308 struct sha1_ctx ctx;
a322f70c 1309
c5afc314
N
1310 if (vcl)
1311 guid = vcl->conf.guid;
1312 else
1313 guid = ddf->anchor.guid;
1314
1315 sha1_init_ctx(&ctx);
1316 sha1_process_bytes(guid, DDF_GUID_LEN, &ctx);
c5afc314
N
1317 sha1_finish_ctx(&ctx, buf);
1318 memcpy(uuid, buf, 4*4);
a322f70c
DW
1319}
1320
a5d85af7 1321static void getinfo_super_ddf_bvd(struct supertype *st, struct mdinfo *info, char *map);
78e44928 1322
a5d85af7 1323static void getinfo_super_ddf(struct supertype *st, struct mdinfo *info, char *map)
a322f70c
DW
1324{
1325 struct ddf_super *ddf = st->sb;
a5d85af7 1326 int map_disks = info->array.raid_disks;
a322f70c 1327
78e44928 1328 if (ddf->currentconf) {
a5d85af7 1329 getinfo_super_ddf_bvd(st, info, map);
78e44928
NB
1330 return;
1331 }
1332
a322f70c
DW
1333 info->array.raid_disks = __be16_to_cpu(ddf->phys->used_pdes);
1334 info->array.level = LEVEL_CONTAINER;
1335 info->array.layout = 0;
1336 info->array.md_minor = -1;
1337 info->array.ctime = DECADE + __be32_to_cpu(*(__u32*)
1338 (ddf->anchor.guid+16));
1339 info->array.utime = 0;
1340 info->array.chunk_size = 0;
510242aa 1341 info->container_enough = 1;
a322f70c 1342
a322f70c
DW
1343
1344 info->disk.major = 0;
1345 info->disk.minor = 0;
cba0191b
NB
1346 if (ddf->dlist) {
1347 info->disk.number = __be32_to_cpu(ddf->dlist->disk.refnum);
59e36268 1348 info->disk.raid_disk = find_phys(ddf, ddf->dlist->disk.refnum);
d2ca6449
NB
1349
1350 info->data_offset = __be64_to_cpu(ddf->phys->
1351 entries[info->disk.raid_disk].
1352 config_size);
1353 info->component_size = ddf->dlist->size - info->data_offset;
cba0191b
NB
1354 } else {
1355 info->disk.number = -1;
661dce36 1356 info->disk.raid_disk = -1;
cba0191b
NB
1357// info->disk.raid_disk = find refnum in the table and use index;
1358 }
f22385f9 1359 info->disk.state = (1 << MD_DISK_SYNC) | (1 << MD_DISK_ACTIVE);
a19c88b8 1360
d2ca6449 1361
921d9e16 1362 info->recovery_start = MaxSector;
a19c88b8 1363 info->reshape_active = 0;
c5afc314 1364 info->name[0] = 0;
a322f70c 1365
f35f2525
N
1366 info->array.major_version = -1;
1367 info->array.minor_version = -2;
159c3a1a 1368 strcpy(info->text_version, "ddf");
a67dd8cc 1369 info->safe_mode_delay = 0;
159c3a1a 1370
c5afc314 1371 uuid_from_super_ddf(st, info->uuid);
a322f70c 1372
a5d85af7
N
1373 if (map) {
1374 int i;
1375 for (i = 0 ; i < map_disks; i++) {
1376 if (i < info->array.raid_disks &&
1377 (__be16_to_cpu(ddf->phys->entries[i].state) & DDF_Online) &&
1378 !(__be16_to_cpu(ddf->phys->entries[i].state) & DDF_Failed))
1379 map[i] = 1;
1380 else
1381 map[i] = 0;
1382 }
1383 }
a322f70c
DW
1384}
1385
598f0d58
NB
1386static int rlq_to_layout(int rlq, int prl, int raiddisks);
1387
a5d85af7 1388static void getinfo_super_ddf_bvd(struct supertype *st, struct mdinfo *info, char *map)
a322f70c
DW
1389{
1390 struct ddf_super *ddf = st->sb;
d2ca6449
NB
1391 struct vcl *vc = ddf->currentconf;
1392 int cd = ddf->currentdev;
db42fa9b 1393 int j;
8592f29d 1394 struct dl *dl;
a5d85af7 1395 int map_disks = info->array.raid_disks;
a322f70c
DW
1396
1397 /* FIXME this returns BVD info - what if we want SVD ?? */
1398
d2ca6449
NB
1399 info->array.raid_disks = __be16_to_cpu(vc->conf.prim_elmnt_count);
1400 info->array.level = map_num1(ddf_level_num, vc->conf.prl);
1401 info->array.layout = rlq_to_layout(vc->conf.rlq, vc->conf.prl,
598f0d58 1402 info->array.raid_disks);
a322f70c 1403 info->array.md_minor = -1;
d2ca6449
NB
1404 info->array.ctime = DECADE +
1405 __be32_to_cpu(*(__u32*)(vc->conf.guid+16));
1406 info->array.utime = DECADE + __be32_to_cpu(vc->conf.timestamp);
1407 info->array.chunk_size = 512 << vc->conf.chunk_shift;
da9b4a62 1408 info->custom_array_size = 0;
d2ca6449 1409
f21e18ca 1410 if (cd >= 0 && (unsigned)cd < ddf->mppe) {
d2ca6449
NB
1411 info->data_offset = __be64_to_cpu(vc->lba_offset[cd]);
1412 if (vc->block_sizes)
1413 info->component_size = vc->block_sizes[cd];
1414 else
1415 info->component_size = __be64_to_cpu(vc->conf.blocks);
1416 }
a322f70c 1417
8592f29d
N
1418 for (dl = ddf->dlist; dl ; dl = dl->next)
1419 if (dl->raiddisk == info->disk.raid_disk)
1420 break;
a322f70c
DW
1421 info->disk.major = 0;
1422 info->disk.minor = 0;
8592f29d
N
1423 if (dl) {
1424 info->disk.major = dl->major;
1425 info->disk.minor = dl->minor;
1426 }
a322f70c
DW
1427// info->disk.number = __be32_to_cpu(ddf->disk.refnum);
1428// info->disk.raid_disk = find refnum in the table and use index;
1429// info->disk.state = ???;
1430
103f2410
NB
1431 info->container_member = ddf->currentconf->vcnum;
1432
921d9e16 1433 info->recovery_start = MaxSector;
80d26cb2 1434 info->resync_start = 0;
624c5ad4 1435 info->reshape_active = 0;
80d26cb2
NB
1436 if (!(ddf->virt->entries[info->container_member].state
1437 & DDF_state_inconsistent) &&
1438 (ddf->virt->entries[info->container_member].init_state
1439 & DDF_initstate_mask)
1440 == DDF_init_full)
b7528a20 1441 info->resync_start = MaxSector;
80d26cb2 1442
a322f70c
DW
1443 uuid_from_super_ddf(st, info->uuid);
1444
f35f2525
N
1445 info->array.major_version = -1;
1446 info->array.minor_version = -2;
9b63e648 1447 sprintf(info->text_version, "/%s/%d",
159c3a1a 1448 devnum2devname(st->container_dev),
9b63e648 1449 info->container_member);
a67dd8cc 1450 info->safe_mode_delay = 200;
159c3a1a 1451
db42fa9b
N
1452 memcpy(info->name, ddf->virt->entries[info->container_member].name, 16);
1453 info->name[16]=0;
1454 for(j=0; j<16; j++)
1455 if (info->name[j] == ' ')
1456 info->name[j] = 0;
a5d85af7
N
1457
1458 if (map)
1459 for (j = 0; j < map_disks; j++) {
1460 map[j] = 0;
1461 if (j < info->array.raid_disks) {
1462 int i = find_phys(ddf, vc->conf.phys_refnum[j]);
1463 if (i >= 0 &&
1464 (__be16_to_cpu(ddf->phys->entries[i].state) & DDF_Online) &&
1465 !(__be16_to_cpu(ddf->phys->entries[i].state) & DDF_Failed))
1466 map[i] = 1;
1467 }
1468 }
a322f70c
DW
1469}
1470
598f0d58 1471
a322f70c
DW
1472static int update_super_ddf(struct supertype *st, struct mdinfo *info,
1473 char *update,
1474 char *devname, int verbose,
1475 int uuid_set, char *homehost)
1476{
1477 /* For 'assemble' and 'force' we need to return non-zero if any
1478 * change was made. For others, the return value is ignored.
1479 * Update options are:
1480 * force-one : This device looks a bit old but needs to be included,
1481 * update age info appropriately.
1482 * assemble: clear any 'faulty' flag to allow this device to
1483 * be assembled.
1484 * force-array: Array is degraded but being forced, mark it clean
1485 * if that will be needed to assemble it.
1486 *
1487 * newdev: not used ????
1488 * grow: Array has gained a new device - this is currently for
1489 * linear only
1490 * resync: mark as dirty so a resync will happen.
59e36268 1491 * uuid: Change the uuid of the array to match what is given
a322f70c
DW
1492 * homehost: update the recorded homehost
1493 * name: update the name - preserving the homehost
1494 * _reshape_progress: record new reshape_progress position.
1495 *
1496 * Following are not relevant for this version:
1497 * sparc2.2 : update from old dodgey metadata
1498 * super-minor: change the preferred_minor number
1499 * summaries: update redundant counters.
1500 */
1501 int rv = 0;
1502// struct ddf_super *ddf = st->sb;
7a7cc504 1503// struct vd_config *vd = find_vdcr(ddf, info->container_member);
a322f70c
DW
1504// struct virtual_entry *ve = find_ve(ddf);
1505
a322f70c
DW
1506 /* we don't need to handle "force-*" or "assemble" as
1507 * there is no need to 'trick' the kernel. We the metadata is
1508 * first updated to activate the array, all the implied modifications
1509 * will just happen.
1510 */
1511
1512 if (strcmp(update, "grow") == 0) {
1513 /* FIXME */
1e2b2765 1514 } else if (strcmp(update, "resync") == 0) {
a322f70c 1515// info->resync_checkpoint = 0;
1e2b2765 1516 } else if (strcmp(update, "homehost") == 0) {
a322f70c
DW
1517 /* homehost is stored in controller->vendor_data,
1518 * or it is when we are the vendor
1519 */
1520// if (info->vendor_is_local)
1521// strcpy(ddf->controller.vendor_data, homehost);
1e2b2765 1522 rv = -1;
f49208ec 1523 } else if (strcmp(update, "name") == 0) {
a322f70c
DW
1524 /* name is stored in virtual_entry->name */
1525// memset(ve->name, ' ', 16);
1526// strncpy(ve->name, info->name, 16);
1e2b2765 1527 rv = -1;
f49208ec 1528 } else if (strcmp(update, "_reshape_progress") == 0) {
a322f70c 1529 /* We don't support reshape yet */
f49208ec
N
1530 } else if (strcmp(update, "assemble") == 0 ) {
1531 /* Do nothing, just succeed */
1532 rv = 0;
1e2b2765
N
1533 } else
1534 rv = -1;
a322f70c
DW
1535
1536// update_all_csum(ddf);
1537
1538 return rv;
1539}
1540
5f8097be
NB
1541static void make_header_guid(char *guid)
1542{
1543 __u32 stamp;
5f8097be
NB
1544 /* Create a DDF Header of Virtual Disk GUID */
1545
1546 /* 24 bytes of fiction required.
1547 * first 8 are a 'vendor-id' - "Linux-MD"
1548 * next 8 are controller type.. how about 0X DEAD BEEF 0000 0000
1549 * Remaining 8 random number plus timestamp
1550 */
1551 memcpy(guid, T10, sizeof(T10));
1552 stamp = __cpu_to_be32(0xdeadbeef);
1553 memcpy(guid+8, &stamp, 4);
1554 stamp = __cpu_to_be32(0);
1555 memcpy(guid+12, &stamp, 4);
1556 stamp = __cpu_to_be32(time(0) - DECADE);
1557 memcpy(guid+16, &stamp, 4);
bfb7ea78 1558 stamp = random32();
5f8097be 1559 memcpy(guid+20, &stamp, 4);
5f8097be 1560}
59e36268 1561
78e44928
NB
1562static int init_super_ddf_bvd(struct supertype *st,
1563 mdu_array_info_t *info,
1564 unsigned long long size,
1565 char *name, char *homehost,
1566 int *uuid);
1567
a322f70c
DW
1568static int init_super_ddf(struct supertype *st,
1569 mdu_array_info_t *info,
1570 unsigned long long size, char *name, char *homehost,
1571 int *uuid)
1572{
1573 /* This is primarily called by Create when creating a new array.
1574 * We will then get add_to_super called for each component, and then
1575 * write_init_super called to write it out to each device.
1576 * For DDF, Create can create on fresh devices or on a pre-existing
1577 * array.
1578 * To create on a pre-existing array a different method will be called.
1579 * This one is just for fresh drives.
1580 *
1581 * We need to create the entire 'ddf' structure which includes:
1582 * DDF headers - these are easy.
1583 * Controller data - a Sector describing this controller .. not that
1584 * this is a controller exactly.
1585 * Physical Disk Record - one entry per device, so
1586 * leave plenty of space.
1587 * Virtual Disk Records - again, just leave plenty of space.
1588 * This just lists VDs, doesn't give details
1589 * Config records - describes the VDs that use this disk
1590 * DiskData - describes 'this' device.
1591 * BadBlockManagement - empty
1592 * Diag Space - empty
1593 * Vendor Logs - Could we put bitmaps here?
1594 *
1595 */
1596 struct ddf_super *ddf;
1597 char hostname[17];
1598 int hostlen;
a322f70c
DW
1599 int max_phys_disks, max_virt_disks;
1600 unsigned long long sector;
1601 int clen;
1602 int i;
1603 int pdsize, vdsize;
1604 struct phys_disk *pd;
1605 struct virtual_disk *vd;
1606
78e44928 1607 if (st->sb)
955e9ea1 1608 return init_super_ddf_bvd(st, info, size, name, homehost, uuid);
ba7eb04f 1609
3d2c4fc7
DW
1610 if (posix_memalign((void**)&ddf, 512, sizeof(*ddf)) != 0) {
1611 fprintf(stderr, Name ": %s could not allocate superblock\n", __func__);
1612 return 0;
1613 }
6264b437 1614 memset(ddf, 0, sizeof(*ddf));
a322f70c
DW
1615 ddf->dlist = NULL; /* no physical disks yet */
1616 ddf->conflist = NULL; /* No virtual disks yet */
955e9ea1
DW
1617 st->sb = ddf;
1618
1619 if (info == NULL) {
1620 /* zeroing superblock */
1621 return 0;
1622 }
a322f70c
DW
1623
1624 /* At least 32MB *must* be reserved for the ddf. So let's just
1625 * start 32MB from the end, and put the primary header there.
1626 * Don't do secondary for now.
1627 * We don't know exactly where that will be yet as it could be
1628 * different on each device. To just set up the lengths.
1629 *
1630 */
1631
1632 ddf->anchor.magic = DDF_HEADER_MAGIC;
5f8097be 1633 make_header_guid(ddf->anchor.guid);
a322f70c 1634
59e36268 1635 memcpy(ddf->anchor.revision, DDF_REVISION_2, 8);
a322f70c
DW
1636 ddf->anchor.seq = __cpu_to_be32(1);
1637 ddf->anchor.timestamp = __cpu_to_be32(time(0) - DECADE);
1638 ddf->anchor.openflag = 0xFF;
1639 ddf->anchor.foreignflag = 0;
1640 ddf->anchor.enforcegroups = 0; /* Is this best?? */
1641 ddf->anchor.pad0 = 0xff;
1642 memset(ddf->anchor.pad1, 0xff, 12);
1643 memset(ddf->anchor.header_ext, 0xff, 32);
1644 ddf->anchor.primary_lba = ~(__u64)0;
1645 ddf->anchor.secondary_lba = ~(__u64)0;
1646 ddf->anchor.type = DDF_HEADER_ANCHOR;
1647 memset(ddf->anchor.pad2, 0xff, 3);
1648 ddf->anchor.workspace_len = __cpu_to_be32(32768); /* Must be reserved */
1649 ddf->anchor.workspace_lba = ~(__u64)0; /* Put this at bottom
1650 of 32M reserved.. */
1651 max_phys_disks = 1023; /* Should be enough */
1652 ddf->anchor.max_pd_entries = __cpu_to_be16(max_phys_disks);
1653 max_virt_disks = 255;
1654 ddf->anchor.max_vd_entries = __cpu_to_be16(max_virt_disks); /* ?? */
1655 ddf->anchor.max_partitions = __cpu_to_be16(64); /* ?? */
1656 ddf->max_part = 64;
8c3b8c2c 1657 ddf->mppe = 256;
59e36268
NB
1658 ddf->conf_rec_len = 1 + ROUND_UP(ddf->mppe * (4+8), 512)/512;
1659 ddf->anchor.config_record_len = __cpu_to_be16(ddf->conf_rec_len);
1660 ddf->anchor.max_primary_element_entries = __cpu_to_be16(ddf->mppe);
a322f70c 1661 memset(ddf->anchor.pad3, 0xff, 54);
a322f70c
DW
1662 /* controller sections is one sector long immediately
1663 * after the ddf header */
1664 sector = 1;
1665 ddf->anchor.controller_section_offset = __cpu_to_be32(sector);
1666 ddf->anchor.controller_section_length = __cpu_to_be32(1);
1667 sector += 1;
1668
1669 /* phys is 8 sectors after that */
1670 pdsize = ROUND_UP(sizeof(struct phys_disk) +
1671 sizeof(struct phys_disk_entry)*max_phys_disks,
1672 512);
1673 switch(pdsize/512) {
1674 case 2: case 8: case 32: case 128: case 512: break;
1675 default: abort();
1676 }
1677 ddf->anchor.phys_section_offset = __cpu_to_be32(sector);
1678 ddf->anchor.phys_section_length =
1679 __cpu_to_be32(pdsize/512); /* max_primary_element_entries/8 */
1680 sector += pdsize/512;
1681
1682 /* virt is another 32 sectors */
1683 vdsize = ROUND_UP(sizeof(struct virtual_disk) +
1684 sizeof(struct virtual_entry) * max_virt_disks,
1685 512);
1686 switch(vdsize/512) {
1687 case 2: case 8: case 32: case 128: case 512: break;
1688 default: abort();
1689 }
1690 ddf->anchor.virt_section_offset = __cpu_to_be32(sector);
1691 ddf->anchor.virt_section_length =
1692 __cpu_to_be32(vdsize/512); /* max_vd_entries/8 */
1693 sector += vdsize/512;
1694
59e36268 1695 clen = ddf->conf_rec_len * (ddf->max_part+1);
a322f70c
DW
1696 ddf->anchor.config_section_offset = __cpu_to_be32(sector);
1697 ddf->anchor.config_section_length = __cpu_to_be32(clen);
1698 sector += clen;
1699
1700 ddf->anchor.data_section_offset = __cpu_to_be32(sector);
1701 ddf->anchor.data_section_length = __cpu_to_be32(1);
1702 sector += 1;
1703
1704 ddf->anchor.bbm_section_length = __cpu_to_be32(0);
1705 ddf->anchor.bbm_section_offset = __cpu_to_be32(0xFFFFFFFF);
1706 ddf->anchor.diag_space_length = __cpu_to_be32(0);
1707 ddf->anchor.diag_space_offset = __cpu_to_be32(0xFFFFFFFF);
1708 ddf->anchor.vendor_length = __cpu_to_be32(0);
1709 ddf->anchor.vendor_offset = __cpu_to_be32(0xFFFFFFFF);
1710
1711 memset(ddf->anchor.pad4, 0xff, 256);
1712
1713 memcpy(&ddf->primary, &ddf->anchor, 512);
1714 memcpy(&ddf->secondary, &ddf->anchor, 512);
1715
1716 ddf->primary.openflag = 1; /* I guess.. */
1717 ddf->primary.type = DDF_HEADER_PRIMARY;
1718
1719 ddf->secondary.openflag = 1; /* I guess.. */
1720 ddf->secondary.type = DDF_HEADER_SECONDARY;
1721
1722 ddf->active = &ddf->primary;
1723
1724 ddf->controller.magic = DDF_CONTROLLER_MAGIC;
1725
1726 /* 24 more bytes of fiction required.
1727 * first 8 are a 'vendor-id' - "Linux-MD"
1728 * Remaining 16 are serial number.... maybe a hostname would do?
1729 */
1730 memcpy(ddf->controller.guid, T10, sizeof(T10));
1ba6bff9
DW
1731 gethostname(hostname, sizeof(hostname));
1732 hostname[sizeof(hostname) - 1] = 0;
a322f70c
DW
1733 hostlen = strlen(hostname);
1734 memcpy(ddf->controller.guid + 24 - hostlen, hostname, hostlen);
1735 for (i = strlen(T10) ; i+hostlen < 24; i++)
1736 ddf->controller.guid[i] = ' ';
1737
1738 ddf->controller.type.vendor_id = __cpu_to_be16(0xDEAD);
1739 ddf->controller.type.device_id = __cpu_to_be16(0xBEEF);
1740 ddf->controller.type.sub_vendor_id = 0;
1741 ddf->controller.type.sub_device_id = 0;
1742 memcpy(ddf->controller.product_id, "What Is My PID??", 16);
1743 memset(ddf->controller.pad, 0xff, 8);
1744 memset(ddf->controller.vendor_data, 0xff, 448);
a9e1c11d
N
1745 if (homehost && strlen(homehost) < 440)
1746 strcpy((char*)ddf->controller.vendor_data, homehost);
a322f70c 1747
3d2c4fc7
DW
1748 if (posix_memalign((void**)&pd, 512, pdsize) != 0) {
1749 fprintf(stderr, Name ": %s could not allocate pd\n", __func__);
1750 return 0;
1751 }
6416d527 1752 ddf->phys = pd;
a322f70c
DW
1753 ddf->pdsize = pdsize;
1754
1755 memset(pd, 0xff, pdsize);
1756 memset(pd, 0, sizeof(*pd));
076515ba 1757 pd->magic = DDF_PHYS_RECORDS_MAGIC;
a322f70c
DW
1758 pd->used_pdes = __cpu_to_be16(0);
1759 pd->max_pdes = __cpu_to_be16(max_phys_disks);
1760 memset(pd->pad, 0xff, 52);
1761
3d2c4fc7
DW
1762 if (posix_memalign((void**)&vd, 512, vdsize) != 0) {
1763 fprintf(stderr, Name ": %s could not allocate vd\n", __func__);
1764 return 0;
1765 }
6416d527 1766 ddf->virt = vd;
a322f70c
DW
1767 ddf->vdsize = vdsize;
1768 memset(vd, 0, vdsize);
1769 vd->magic = DDF_VIRT_RECORDS_MAGIC;
1770 vd->populated_vdes = __cpu_to_be16(0);
1771 vd->max_vdes = __cpu_to_be16(max_virt_disks);
1772 memset(vd->pad, 0xff, 52);
1773
5f8097be
NB
1774 for (i=0; i<max_virt_disks; i++)
1775 memset(&vd->entries[i], 0xff, sizeof(struct virtual_entry));
1776
a322f70c 1777 st->sb = ddf;
18a2f463 1778 ddf->updates_pending = 1;
a322f70c
DW
1779 return 1;
1780}
1781
5f8097be
NB
1782static int chunk_to_shift(int chunksize)
1783{
1784 return ffs(chunksize/512)-1;
1785}
1786
1787static int level_to_prl(int level)
1788{
1789 switch (level) {
1790 case LEVEL_LINEAR: return DDF_CONCAT;
1791 case 0: return DDF_RAID0;
1792 case 1: return DDF_RAID1;
1793 case 4: return DDF_RAID4;
1794 case 5: return DDF_RAID5;
1795 case 6: return DDF_RAID6;
1796 default: return -1;
1797 }
1798}
1799static int layout_to_rlq(int level, int layout, int raiddisks)
1800{
1801 switch(level) {
1802 case 0:
1803 return DDF_RAID0_SIMPLE;
1804 case 1:
1805 switch(raiddisks) {
1806 case 2: return DDF_RAID1_SIMPLE;
1807 case 3: return DDF_RAID1_MULTI;
1808 default: return -1;
1809 }
1810 case 4:
1811 switch(layout) {
1812 case 0: return DDF_RAID4_N;
1813 }
1814 break;
1815 case 5:
5f8097be
NB
1816 switch(layout) {
1817 case ALGORITHM_LEFT_ASYMMETRIC:
1818 return DDF_RAID5_N_RESTART;
1819 case ALGORITHM_RIGHT_ASYMMETRIC:
b640a252 1820 return DDF_RAID5_0_RESTART;
5f8097be
NB
1821 case ALGORITHM_LEFT_SYMMETRIC:
1822 return DDF_RAID5_N_CONTINUE;
1823 case ALGORITHM_RIGHT_SYMMETRIC:
1824 return -1; /* not mentioned in standard */
1825 }
b640a252
N
1826 case 6:
1827 switch(layout) {
1828 case ALGORITHM_ROTATING_N_RESTART:
1829 return DDF_RAID5_N_RESTART;
1830 case ALGORITHM_ROTATING_ZERO_RESTART:
1831 return DDF_RAID6_0_RESTART;
1832 case ALGORITHM_ROTATING_N_CONTINUE:
1833 return DDF_RAID5_N_CONTINUE;
1834 }
5f8097be
NB
1835 }
1836 return -1;
1837}
1838
598f0d58
NB
1839static int rlq_to_layout(int rlq, int prl, int raiddisks)
1840{
1841 switch(prl) {
1842 case DDF_RAID0:
1843 return 0; /* hopefully rlq == DDF_RAID0_SIMPLE */
1844 case DDF_RAID1:
1845 return 0; /* hopefully rlq == SIMPLE or MULTI depending
1846 on raiddisks*/
1847 case DDF_RAID4:
1848 switch(rlq) {
1849 case DDF_RAID4_N:
1850 return 0;
1851 default:
1852 /* not supported */
1853 return -1; /* FIXME this isn't checked */
1854 }
1855 case DDF_RAID5:
598f0d58
NB
1856 switch(rlq) {
1857 case DDF_RAID5_N_RESTART:
1858 return ALGORITHM_LEFT_ASYMMETRIC;
1859 case DDF_RAID5_0_RESTART:
1860 return ALGORITHM_RIGHT_ASYMMETRIC;
1861 case DDF_RAID5_N_CONTINUE:
1862 return ALGORITHM_LEFT_SYMMETRIC;
1863 default:
1864 return -1;
1865 }
59e36268
NB
1866 case DDF_RAID6:
1867 switch(rlq) {
1868 case DDF_RAID5_N_RESTART:
b640a252 1869 return ALGORITHM_ROTATING_N_RESTART;
59e36268 1870 case DDF_RAID6_0_RESTART:
b640a252 1871 return ALGORITHM_ROTATING_ZERO_RESTART;
59e36268 1872 case DDF_RAID5_N_CONTINUE:
b640a252 1873 return ALGORITHM_ROTATING_N_CONTINUE;
59e36268
NB
1874 default:
1875 return -1;
1876 }
598f0d58
NB
1877 }
1878 return -1;
1879}
1880
0e600426 1881#ifndef MDASSEMBLE
59e36268
NB
1882struct extent {
1883 unsigned long long start, size;
1884};
78e44928 1885static int cmp_extent(const void *av, const void *bv)
59e36268
NB
1886{
1887 const struct extent *a = av;
1888 const struct extent *b = bv;
1889 if (a->start < b->start)
1890 return -1;
1891 if (a->start > b->start)
1892 return 1;
1893 return 0;
1894}
1895
78e44928 1896static struct extent *get_extents(struct ddf_super *ddf, struct dl *dl)
59e36268
NB
1897{
1898 /* find a list of used extents on the give physical device
1899 * (dnum) of the given ddf.
1900 * Return a malloced array of 'struct extent'
1901
1902FIXME ignore DDF_Legacy devices?
1903
1904 */
1905 struct extent *rv;
1906 int n = 0;
f21e18ca 1907 unsigned int i, j;
59e36268
NB
1908
1909 rv = malloc(sizeof(struct extent) * (ddf->max_part + 2));
1910 if (!rv)
1911 return NULL;
1912
1913 for (i = 0; i < ddf->max_part; i++) {
1914 struct vcl *v = dl->vlist[i];
1915 if (v == NULL)
1916 continue;
f21e18ca 1917 for (j = 0; j < v->conf.prim_elmnt_count; j++)
59e36268
NB
1918 if (v->conf.phys_refnum[j] == dl->disk.refnum) {
1919 /* This device plays role 'j' in 'v'. */
1920 rv[n].start = __be64_to_cpu(v->lba_offset[j]);
1921 rv[n].size = __be64_to_cpu(v->conf.blocks);
1922 n++;
1923 break;
1924 }
1925 }
1926 qsort(rv, n, sizeof(*rv), cmp_extent);
1927
1928 rv[n].start = __be64_to_cpu(ddf->phys->entries[dl->pdnum].config_size);
1929 rv[n].size = 0;
1930 return rv;
1931}
0e600426 1932#endif
59e36268 1933
5f8097be
NB
1934static int init_super_ddf_bvd(struct supertype *st,
1935 mdu_array_info_t *info,
1936 unsigned long long size,
1937 char *name, char *homehost,
1938 int *uuid)
1939{
1940 /* We are creating a BVD inside a pre-existing container.
1941 * so st->sb is already set.
1942 * We need to create a new vd_config and a new virtual_entry
1943 */
1944 struct ddf_super *ddf = st->sb;
f21e18ca 1945 unsigned int venum;
5f8097be
NB
1946 struct virtual_entry *ve;
1947 struct vcl *vcl;
1948 struct vd_config *vc;
5f8097be
NB
1949
1950 if (__be16_to_cpu(ddf->virt->populated_vdes)
1951 >= __be16_to_cpu(ddf->virt->max_vdes)) {
1952 fprintf(stderr, Name": This ddf already has the "
1953 "maximum of %d virtual devices\n",
1954 __be16_to_cpu(ddf->virt->max_vdes));
1955 return 0;
1956 }
1957
97c9c100
N
1958 if (name)
1959 for (venum = 0; venum < __be16_to_cpu(ddf->virt->max_vdes); venum++)
1960 if (!all_ff(ddf->virt->entries[venum].guid)) {
1961 char *n = ddf->virt->entries[venum].name;
1962
1963 if (strncmp(name, n, 16) == 0) {
1964 fprintf(stderr, Name ": This ddf already"
1965 " has an array called %s\n",
1966 name);
1967 return 0;
1968 }
1969 }
1970
5f8097be
NB
1971 for (venum = 0; venum < __be16_to_cpu(ddf->virt->max_vdes); venum++)
1972 if (all_ff(ddf->virt->entries[venum].guid))
1973 break;
1974 if (venum == __be16_to_cpu(ddf->virt->max_vdes)) {
1975 fprintf(stderr, Name ": Cannot find spare slot for "
1976 "virtual disk - DDF is corrupt\n");
1977 return 0;
1978 }
1979 ve = &ddf->virt->entries[venum];
1980
1981 /* A Virtual Disk GUID contains the T10 Vendor ID, controller type,
1982 * timestamp, random number
1983 */
1984 make_header_guid(ve->guid);
1985 ve->unit = __cpu_to_be16(info->md_minor);
1986 ve->pad0 = 0xFFFF;
1987 ve->guid_crc = crc32(0, (unsigned char*)ddf->anchor.guid, DDF_GUID_LEN);
1988 ve->type = 0;
7a7cc504
NB
1989 ve->state = DDF_state_degraded; /* Will be modified as devices are added */
1990 if (info->state & 1) /* clean */
1991 ve->init_state = DDF_init_full;
1992 else
1993 ve->init_state = DDF_init_not;
1994
5f8097be
NB
1995 memset(ve->pad1, 0xff, 14);
1996 memset(ve->name, ' ', 16);
1997 if (name)
1998 strncpy(ve->name, name, 16);
1999 ddf->virt->populated_vdes =
2000 __cpu_to_be16(__be16_to_cpu(ddf->virt->populated_vdes)+1);
2001
2002 /* Now create a new vd_config */
3d2c4fc7
DW
2003 if (posix_memalign((void**)&vcl, 512,
2004 (offsetof(struct vcl, conf) + ddf->conf_rec_len * 512)) != 0) {
2005 fprintf(stderr, Name ": %s could not allocate vd_config\n", __func__);
2006 return 0;
2007 }
8c3b8c2c 2008 vcl->lba_offset = (__u64*) &vcl->conf.phys_refnum[ddf->mppe];
59e36268
NB
2009 vcl->vcnum = venum;
2010 vcl->block_sizes = NULL; /* FIXME not for CONCAT */
5f8097be
NB
2011
2012 vc = &vcl->conf;
2013
2014 vc->magic = DDF_VD_CONF_MAGIC;
2015 memcpy(vc->guid, ve->guid, DDF_GUID_LEN);
2016 vc->timestamp = __cpu_to_be32(time(0)-DECADE);
2017 vc->seqnum = __cpu_to_be32(1);
2018 memset(vc->pad0, 0xff, 24);
2019 vc->prim_elmnt_count = __cpu_to_be16(info->raid_disks);
2020 vc->chunk_shift = chunk_to_shift(info->chunk_size);
2021 vc->prl = level_to_prl(info->level);
2022 vc->rlq = layout_to_rlq(info->level, info->layout, info->raid_disks);
2023 vc->sec_elmnt_count = 1;
2024 vc->sec_elmnt_seq = 0;
2025 vc->srl = 0;
2026 vc->blocks = __cpu_to_be64(info->size * 2);
2027 vc->array_blocks = __cpu_to_be64(
2028 calc_array_size(info->level, info->raid_disks, info->layout,
2029 info->chunk_size, info->size*2));
2030 memset(vc->pad1, 0xff, 8);
2031 vc->spare_refs[0] = 0xffffffff;
2032 vc->spare_refs[1] = 0xffffffff;
2033 vc->spare_refs[2] = 0xffffffff;
2034 vc->spare_refs[3] = 0xffffffff;
2035 vc->spare_refs[4] = 0xffffffff;
2036 vc->spare_refs[5] = 0xffffffff;
2037 vc->spare_refs[6] = 0xffffffff;
2038 vc->spare_refs[7] = 0xffffffff;
2039 memset(vc->cache_pol, 0, 8);
2040 vc->bg_rate = 0x80;
2041 memset(vc->pad2, 0xff, 3);
2042 memset(vc->pad3, 0xff, 52);
2043 memset(vc->pad4, 0xff, 192);
2044 memset(vc->v0, 0xff, 32);
2045 memset(vc->v1, 0xff, 32);
2046 memset(vc->v2, 0xff, 16);
2047 memset(vc->v3, 0xff, 16);
2048 memset(vc->vendor, 0xff, 32);
598f0d58 2049
8c3b8c2c 2050 memset(vc->phys_refnum, 0xff, 4*ddf->mppe);
e5a2a3cf 2051 memset(vc->phys_refnum+ddf->mppe, 0x00, 8*ddf->mppe);
5f8097be
NB
2052
2053 vcl->next = ddf->conflist;
2054 ddf->conflist = vcl;
d2ca6449 2055 ddf->currentconf = vcl;
18a2f463 2056 ddf->updates_pending = 1;
5f8097be
NB
2057 return 1;
2058}
2059
0e600426 2060#ifndef MDASSEMBLE
5f8097be
NB
2061static void add_to_super_ddf_bvd(struct supertype *st,
2062 mdu_disk_info_t *dk, int fd, char *devname)
2063{
2064 /* fd and devname identify a device with-in the ddf container (st).
2065 * dk identifies a location in the new BVD.
2066 * We need to find suitable free space in that device and update
2067 * the phys_refnum and lba_offset for the newly created vd_config.
2068 * We might also want to update the type in the phys_disk
5575e7d9 2069 * section.
8592f29d
N
2070 *
2071 * Alternately: fd == -1 and we have already chosen which device to
2072 * use and recorded in dlist->raid_disk;
5f8097be
NB
2073 */
2074 struct dl *dl;
2075 struct ddf_super *ddf = st->sb;
2076 struct vd_config *vc;
2077 __u64 *lba_offset;
f21e18ca
N
2078 unsigned int working;
2079 unsigned int i;
59e36268
NB
2080 unsigned long long blocks, pos, esize;
2081 struct extent *ex;
5f8097be 2082
8592f29d
N
2083 if (fd == -1) {
2084 for (dl = ddf->dlist; dl ; dl = dl->next)
2085 if (dl->raiddisk == dk->raid_disk)
2086 break;
2087 } else {
2088 for (dl = ddf->dlist; dl ; dl = dl->next)
2089 if (dl->major == dk->major &&
2090 dl->minor == dk->minor)
2091 break;
2092 }
5f8097be
NB
2093 if (!dl || ! (dk->state & (1<<MD_DISK_SYNC)))
2094 return;
2095
d2ca6449
NB
2096 vc = &ddf->currentconf->conf;
2097 lba_offset = ddf->currentconf->lba_offset;
59e36268
NB
2098
2099 ex = get_extents(ddf, dl);
2100 if (!ex)
2101 return;
2102
2103 i = 0; pos = 0;
2104 blocks = __be64_to_cpu(vc->blocks);
d2ca6449
NB
2105 if (ddf->currentconf->block_sizes)
2106 blocks = ddf->currentconf->block_sizes[dk->raid_disk];
59e36268
NB
2107
2108 do {
2109 esize = ex[i].start - pos;
2110 if (esize >= blocks)
2111 break;
2112 pos = ex[i].start + ex[i].size;
2113 i++;
2114 } while (ex[i-1].size);
2115
2116 free(ex);
2117 if (esize < blocks)
2118 return;
2119
d2ca6449 2120 ddf->currentdev = dk->raid_disk;
5f8097be 2121 vc->phys_refnum[dk->raid_disk] = dl->disk.refnum;
59e36268 2122 lba_offset[dk->raid_disk] = __cpu_to_be64(pos);
5f8097be 2123
f21e18ca 2124 for (i = 0; i < ddf->max_part ; i++)
5575e7d9
NB
2125 if (dl->vlist[i] == NULL)
2126 break;
2127 if (i == ddf->max_part)
2128 return;
d2ca6449 2129 dl->vlist[i] = ddf->currentconf;
5f8097be 2130
8592f29d
N
2131 if (fd >= 0)
2132 dl->fd = fd;
2133 if (devname)
2134 dl->devname = devname;
7a7cc504
NB
2135
2136 /* Check how many working raid_disks, and if we can mark
2137 * array as optimal yet
2138 */
2139 working = 0;
5575e7d9 2140
f21e18ca 2141 for (i = 0; i < __be16_to_cpu(vc->prim_elmnt_count); i++)
7a7cc504
NB
2142 if (vc->phys_refnum[i] != 0xffffffff)
2143 working++;
59e36268 2144
5575e7d9 2145 /* Find which virtual_entry */
d2ca6449 2146 i = ddf->currentconf->vcnum;
7a7cc504 2147 if (working == __be16_to_cpu(vc->prim_elmnt_count))
5575e7d9
NB
2148 ddf->virt->entries[i].state =
2149 (ddf->virt->entries[i].state & ~DDF_state_mask)
7a7cc504
NB
2150 | DDF_state_optimal;
2151
2152 if (vc->prl == DDF_RAID6 &&
2153 working+1 == __be16_to_cpu(vc->prim_elmnt_count))
5575e7d9
NB
2154 ddf->virt->entries[i].state =
2155 (ddf->virt->entries[i].state & ~DDF_state_mask)
7a7cc504 2156 | DDF_state_part_optimal;
5575e7d9
NB
2157
2158 ddf->phys->entries[dl->pdnum].type &= ~__cpu_to_be16(DDF_Global_Spare);
2159 ddf->phys->entries[dl->pdnum].type |= __cpu_to_be16(DDF_Active_in_VD);
18a2f463 2160 ddf->updates_pending = 1;
5f8097be
NB
2161}
2162
a322f70c
DW
2163/* add a device to a container, either while creating it or while
2164 * expanding a pre-existing container
2165 */
f20c3968 2166static int add_to_super_ddf(struct supertype *st,
a322f70c
DW
2167 mdu_disk_info_t *dk, int fd, char *devname)
2168{
2169 struct ddf_super *ddf = st->sb;
2170 struct dl *dd;
2171 time_t now;
2172 struct tm *tm;
2173 unsigned long long size;
2174 struct phys_disk_entry *pde;
f21e18ca 2175 unsigned int n, i;
a322f70c
DW
2176 struct stat stb;
2177
78e44928
NB
2178 if (ddf->currentconf) {
2179 add_to_super_ddf_bvd(st, dk, fd, devname);
f20c3968 2180 return 0;
78e44928
NB
2181 }
2182
a322f70c
DW
2183 /* This is device numbered dk->number. We need to create
2184 * a phys_disk entry and a more detailed disk_data entry.
2185 */
2186 fstat(fd, &stb);
3d2c4fc7
DW
2187 if (posix_memalign((void**)&dd, 512,
2188 sizeof(*dd) + sizeof(dd->vlist[0]) * ddf->max_part) != 0) {
2189 fprintf(stderr, Name
2190 ": %s could allocate buffer for new disk, aborting\n",
2191 __func__);
f20c3968 2192 return 1;
3d2c4fc7 2193 }
a322f70c
DW
2194 dd->major = major(stb.st_rdev);
2195 dd->minor = minor(stb.st_rdev);
2196 dd->devname = devname;
a322f70c 2197 dd->fd = fd;
b2280677 2198 dd->spare = NULL;
a322f70c
DW
2199
2200 dd->disk.magic = DDF_PHYS_DATA_MAGIC;
2201 now = time(0);
2202 tm = localtime(&now);
2203 sprintf(dd->disk.guid, "%8s%04d%02d%02d",
2204 T10, tm->tm_year+1900, tm->tm_mon+1, tm->tm_mday);
bfb7ea78
N
2205 *(__u32*)(dd->disk.guid + 16) = random32();
2206 *(__u32*)(dd->disk.guid + 20) = random32();
a322f70c 2207
59e36268
NB
2208 do {
2209 /* Cannot be bothered finding a CRC of some irrelevant details*/
bfb7ea78 2210 dd->disk.refnum = random32();
f21e18ca
N
2211 for (i = __be16_to_cpu(ddf->active->max_pd_entries);
2212 i > 0; i--)
2213 if (ddf->phys->entries[i-1].refnum == dd->disk.refnum)
59e36268 2214 break;
f21e18ca 2215 } while (i > 0);
59e36268 2216
a322f70c
DW
2217 dd->disk.forced_ref = 1;
2218 dd->disk.forced_guid = 1;
2219 memset(dd->disk.vendor, ' ', 32);
2220 memcpy(dd->disk.vendor, "Linux", 5);
2221 memset(dd->disk.pad, 0xff, 442);
b2280677 2222 for (i = 0; i < ddf->max_part ; i++)
a322f70c
DW
2223 dd->vlist[i] = NULL;
2224
2225 n = __be16_to_cpu(ddf->phys->used_pdes);
2226 pde = &ddf->phys->entries[n];
5575e7d9
NB
2227 dd->pdnum = n;
2228
2cc2983d
N
2229 if (st->update_tail) {
2230 int len = (sizeof(struct phys_disk) +
2231 sizeof(struct phys_disk_entry));
2232 struct phys_disk *pd;
2233
2234 pd = malloc(len);
2235 pd->magic = DDF_PHYS_RECORDS_MAGIC;
2236 pd->used_pdes = __cpu_to_be16(n);
2237 pde = &pd->entries[0];
2238 dd->mdupdate = pd;
2239 } else {
2240 n++;
2241 ddf->phys->used_pdes = __cpu_to_be16(n);
2242 }
a322f70c
DW
2243
2244 memcpy(pde->guid, dd->disk.guid, DDF_GUID_LEN);
2245 pde->refnum = dd->disk.refnum;
5575e7d9 2246 pde->type = __cpu_to_be16(DDF_Forced_PD_GUID | DDF_Global_Spare);
a322f70c
DW
2247 pde->state = __cpu_to_be16(DDF_Online);
2248 get_dev_size(fd, NULL, &size);
2249 /* We are required to reserve 32Meg, and record the size in sectors */
2250 pde->config_size = __cpu_to_be64( (size - 32*1024*1024) / 512);
2251 sprintf(pde->path, "%17.17s","Information: nil") ;
2252 memset(pde->pad, 0xff, 6);
2253
d2ca6449 2254 dd->size = size >> 9;
2cc2983d
N
2255 if (st->update_tail) {
2256 dd->next = ddf->add_list;
2257 ddf->add_list = dd;
2258 } else {
2259 dd->next = ddf->dlist;
2260 ddf->dlist = dd;
2261 ddf->updates_pending = 1;
2262 }
f20c3968
DW
2263
2264 return 0;
a322f70c
DW
2265}
2266
2267/*
2268 * This is the write_init_super method for a ddf container. It is
2269 * called when creating a container or adding another device to a
2270 * container.
2271 */
2272
6416d527 2273static unsigned char null_conf[4096+512];
18a2f463 2274
1cc7f4fe 2275static int __write_init_super_ddf(struct supertype *st)
a322f70c
DW
2276{
2277
2278 struct ddf_super *ddf = st->sb;
2279 int i;
2280 struct dl *d;
2281 int n_config;
2282 int conf_size;
175593bf
DW
2283 int attempts = 0;
2284 int successes = 0;
a322f70c
DW
2285 unsigned long long size, sector;
2286
175593bf
DW
2287 /* try to write updated metadata,
2288 * if we catch a failure move on to the next disk
2289 */
a322f70c
DW
2290 for (d = ddf->dlist; d; d=d->next) {
2291 int fd = d->fd;
2292
2293 if (fd < 0)
2294 continue;
2295
175593bf 2296 attempts++;
a322f70c
DW
2297 /* We need to fill in the primary, (secondary) and workspace
2298 * lba's in the headers, set their checksums,
2299 * Also checksum phys, virt....
2300 *
2301 * Then write everything out, finally the anchor is written.
2302 */
2303 get_dev_size(fd, NULL, &size);
2304 size /= 512;
2305 ddf->anchor.workspace_lba = __cpu_to_be64(size - 32*1024*2);
2306 ddf->anchor.primary_lba = __cpu_to_be64(size - 16*1024*2);
2307 ddf->anchor.seq = __cpu_to_be32(1);
2308 memcpy(&ddf->primary, &ddf->anchor, 512);
2309 memcpy(&ddf->secondary, &ddf->anchor, 512);
2310
2311 ddf->anchor.openflag = 0xFF; /* 'open' means nothing */
2312 ddf->anchor.seq = 0xFFFFFFFF; /* no sequencing in anchor */
2313 ddf->anchor.crc = calc_crc(&ddf->anchor, 512);
2314
2315 ddf->primary.openflag = 0;
2316 ddf->primary.type = DDF_HEADER_PRIMARY;
2317
2318 ddf->secondary.openflag = 0;
2319 ddf->secondary.type = DDF_HEADER_SECONDARY;
2320
2321 ddf->primary.crc = calc_crc(&ddf->primary, 512);
2322 ddf->secondary.crc = calc_crc(&ddf->secondary, 512);
2323
2324 sector = size - 16*1024*2;
2325 lseek64(fd, sector<<9, 0);
175593bf
DW
2326 if (write(fd, &ddf->primary, 512) < 0)
2327 continue;
a322f70c
DW
2328
2329 ddf->controller.crc = calc_crc(&ddf->controller, 512);
175593bf
DW
2330 if (write(fd, &ddf->controller, 512) < 0)
2331 continue;
a322f70c
DW
2332
2333 ddf->phys->crc = calc_crc(ddf->phys, ddf->pdsize);
2334
175593bf
DW
2335 if (write(fd, ddf->phys, ddf->pdsize) < 0)
2336 continue;
a322f70c
DW
2337
2338 ddf->virt->crc = calc_crc(ddf->virt, ddf->vdsize);
175593bf
DW
2339 if (write(fd, ddf->virt, ddf->vdsize) < 0)
2340 continue;
a322f70c
DW
2341
2342 /* Now write lots of config records. */
8c3b8c2c
NB
2343 n_config = ddf->max_part;
2344 conf_size = ddf->conf_rec_len * 512;
a322f70c
DW
2345 for (i = 0 ; i <= n_config ; i++) {
2346 struct vcl *c = d->vlist[i];
b2280677
NB
2347 if (i == n_config)
2348 c = (struct vcl*)d->spare;
a322f70c
DW
2349
2350 if (c) {
2351 c->conf.crc = calc_crc(&c->conf, conf_size);
175593bf
DW
2352 if (write(fd, &c->conf, conf_size) < 0)
2353 break;
a322f70c 2354 } else {
6416d527 2355 char *null_aligned = (char*)((((unsigned long)null_conf)+511)&~511UL);
18a2f463
NB
2356 if (null_conf[0] != 0xff)
2357 memset(null_conf, 0xff, sizeof(null_conf));
f21e18ca 2358 unsigned int togo = conf_size;
6416d527 2359 while (togo > sizeof(null_conf)-512) {
175593bf
DW
2360 if (write(fd, null_aligned, sizeof(null_conf)-512) < 0)
2361 break;
6416d527 2362 togo -= sizeof(null_conf)-512;
18a2f463 2363 }
175593bf
DW
2364 if (write(fd, null_aligned, togo) < 0)
2365 break;
a322f70c
DW
2366 }
2367 }
175593bf
DW
2368 if (i <= n_config)
2369 continue;
a322f70c 2370 d->disk.crc = calc_crc(&d->disk, 512);
175593bf
DW
2371 if (write(fd, &d->disk, 512) < 0)
2372 continue;
a322f70c
DW
2373
2374 /* Maybe do the same for secondary */
2375
2376 lseek64(fd, (size-1)*512, SEEK_SET);
175593bf
DW
2377 if (write(fd, &ddf->anchor, 512) < 0)
2378 continue;
2379 successes++;
2380 }
2381
175593bf 2382 return attempts != successes;
a322f70c 2383}
7a7cc504
NB
2384
2385static int write_init_super_ddf(struct supertype *st)
2386{
9b1fb677
DW
2387 struct ddf_super *ddf = st->sb;
2388 struct vcl *currentconf = ddf->currentconf;
2389
2390 /* we are done with currentconf reset it to point st at the container */
2391 ddf->currentconf = NULL;
edd8d13c
NB
2392
2393 if (st->update_tail) {
2394 /* queue the virtual_disk and vd_config as metadata updates */
2395 struct virtual_disk *vd;
2396 struct vd_config *vc;
edd8d13c
NB
2397 int len;
2398
9b1fb677 2399 if (!currentconf) {
2cc2983d
N
2400 int len = (sizeof(struct phys_disk) +
2401 sizeof(struct phys_disk_entry));
2402
2403 /* adding a disk to the container. */
2404 if (!ddf->add_list)
2405 return 0;
2406
2407 append_metadata_update(st, ddf->add_list->mdupdate, len);
2408 ddf->add_list->mdupdate = NULL;
2409 return 0;
2410 }
2411
2412 /* Newly created VD */
2413
edd8d13c
NB
2414 /* First the virtual disk. We have a slightly fake header */
2415 len = sizeof(struct virtual_disk) + sizeof(struct virtual_entry);
2416 vd = malloc(len);
2417 *vd = *ddf->virt;
9b1fb677
DW
2418 vd->entries[0] = ddf->virt->entries[currentconf->vcnum];
2419 vd->populated_vdes = __cpu_to_be16(currentconf->vcnum);
edd8d13c
NB
2420 append_metadata_update(st, vd, len);
2421
2422 /* Then the vd_config */
2423 len = ddf->conf_rec_len * 512;
2424 vc = malloc(len);
9b1fb677 2425 memcpy(vc, &currentconf->conf, len);
edd8d13c
NB
2426 append_metadata_update(st, vc, len);
2427
2428 /* FIXME I need to close the fds! */
2429 return 0;
d682f344
N
2430 } else {
2431 struct dl *d;
2432 for (d = ddf->dlist; d; d=d->next)
2433 while (Kill(d->devname, NULL, 0, 1, 1) == 0);
1cc7f4fe 2434 return __write_init_super_ddf(st);
d682f344 2435 }
7a7cc504
NB
2436}
2437
a322f70c
DW
2438#endif
2439
2440static __u64 avail_size_ddf(struct supertype *st, __u64 devsize)
2441{
2442 /* We must reserve the last 32Meg */
2443 if (devsize <= 32*1024*2)
2444 return 0;
2445 return devsize - 32*1024*2;
2446}
2447
2448#ifndef MDASSEMBLE
8592f29d
N
2449
2450static int reserve_space(struct supertype *st, int raiddisks,
2451 unsigned long long size, int chunk,
2452 unsigned long long *freesize)
2453{
2454 /* Find 'raiddisks' spare extents at least 'size' big (but
2455 * only caring about multiples of 'chunk') and remember
2456 * them.
2457 * If the cannot be found, fail.
2458 */
2459 struct dl *dl;
2460 struct ddf_super *ddf = st->sb;
2461 int cnt = 0;
2462
2463 for (dl = ddf->dlist; dl ; dl=dl->next) {
2464 dl->raiddisk = -1;
2465 dl->esize = 0;
2466 }
2467 /* Now find largest extent on each device */
2468 for (dl = ddf->dlist ; dl ; dl=dl->next) {
2469 struct extent *e = get_extents(ddf, dl);
2470 unsigned long long pos = 0;
2471 int i = 0;
2472 int found = 0;
2473 unsigned long long minsize = size;
2474
2475 if (size == 0)
2476 minsize = chunk;
2477
2478 if (!e)
2479 continue;
2480 do {
2481 unsigned long long esize;
2482 esize = e[i].start - pos;
2483 if (esize >= minsize) {
2484 found = 1;
2485 minsize = esize;
2486 }
2487 pos = e[i].start + e[i].size;
2488 i++;
2489 } while (e[i-1].size);
2490 if (found) {
2491 cnt++;
2492 dl->esize = minsize;
2493 }
2494 free(e);
2495 }
2496 if (cnt < raiddisks) {
2497 fprintf(stderr, Name ": not enough devices with space to create array.\n");
2498 return 0; /* No enough free spaces large enough */
2499 }
2500 if (size == 0) {
2501 /* choose the largest size of which there are at least 'raiddisk' */
2502 for (dl = ddf->dlist ; dl ; dl=dl->next) {
2503 struct dl *dl2;
2504 if (dl->esize <= size)
2505 continue;
2506 /* This is bigger than 'size', see if there are enough */
2507 cnt = 0;
2508 for (dl2 = dl; dl2 ; dl2=dl2->next)
2509 if (dl2->esize >= dl->esize)
2510 cnt++;
2511 if (cnt >= raiddisks)
2512 size = dl->esize;
2513 }
2514 if (chunk) {
2515 size = size / chunk;
2516 size *= chunk;
2517 }
2518 *freesize = size;
2519 if (size < 32) {
2520 fprintf(stderr, Name ": not enough spare devices to create array.\n");
2521 return 0;
2522 }
2523 }
2524 /* We have a 'size' of which there are enough spaces.
2525 * We simply do a first-fit */
2526 cnt = 0;
2527 for (dl = ddf->dlist ; dl && cnt < raiddisks ; dl=dl->next) {
2528 if (dl->esize < size)
2529 continue;
2530
2531 dl->raiddisk = cnt;
2532 cnt++;
2533 }
2534 return 1;
2535}
2536
2537
2538
2c514b71
NB
2539static int
2540validate_geometry_ddf_container(struct supertype *st,
2541 int level, int layout, int raiddisks,
2542 int chunk, unsigned long long size,
2543 char *dev, unsigned long long *freesize,
2544 int verbose);
78e44928
NB
2545
2546static int validate_geometry_ddf_bvd(struct supertype *st,
2547 int level, int layout, int raiddisks,
c21e737b 2548 int *chunk, unsigned long long size,
2c514b71
NB
2549 char *dev, unsigned long long *freesize,
2550 int verbose);
78e44928
NB
2551
2552static int validate_geometry_ddf(struct supertype *st,
2c514b71 2553 int level, int layout, int raiddisks,
c21e737b 2554 int *chunk, unsigned long long size,
2c514b71
NB
2555 char *dev, unsigned long long *freesize,
2556 int verbose)
a322f70c
DW
2557{
2558 int fd;
2559 struct mdinfo *sra;
2560 int cfd;
2561
2562 /* ddf potentially supports lots of things, but it depends on
2563 * what devices are offered (and maybe kernel version?)
2564 * If given unused devices, we will make a container.
2565 * If given devices in a container, we will make a BVD.
2566 * If given BVDs, we make an SVD, changing all the GUIDs in the process.
2567 */
2568
bb7295f1
N
2569 if (chunk && *chunk == UnSet)
2570 *chunk = DEFAULT_CHUNK;
2571
2572
a322f70c 2573 if (level == LEVEL_CONTAINER) {
78e44928
NB
2574 /* Must be a fresh device to add to a container */
2575 return validate_geometry_ddf_container(st, level, layout,
c21e737b 2576 raiddisks, chunk?*chunk:0,
2c514b71
NB
2577 size, dev, freesize,
2578 verbose);
5f8097be
NB
2579 }
2580
78e44928
NB
2581 if (!dev) {
2582 /* Initial sanity check. Exclude illegal levels. */
2583 int i;
2584 for (i=0; ddf_level_num[i].num1 != MAXINT; i++)
2585 if (ddf_level_num[i].num2 == level)
2586 break;
b42f577a
N
2587 if (ddf_level_num[i].num1 == MAXINT) {
2588 if (verbose)
2589 fprintf(stderr, Name ": DDF does not support level %d arrays\n",
2590 level);
78e44928 2591 return 0;
b42f577a 2592 }
78e44928 2593 /* Should check layout? etc */
8592f29d
N
2594
2595 if (st->sb && freesize) {
2596 /* --create was given a container to create in.
2597 * So we need to check that there are enough
2598 * free spaces and return the amount of space.
2599 * We may as well remember which drives were
2600 * chosen so that add_to_super/getinfo_super
2601 * can return them.
2602 */
c21e737b 2603 return reserve_space(st, raiddisks, size, chunk?*chunk:0, freesize);
8592f29d 2604 }
a322f70c 2605 return 1;
78e44928 2606 }
a322f70c 2607
8592f29d
N
2608 if (st->sb) {
2609 /* A container has already been opened, so we are
2610 * creating in there. Maybe a BVD, maybe an SVD.
2611 * Should make a distinction one day.
2612 */
2613 return validate_geometry_ddf_bvd(st, level, layout, raiddisks,
2614 chunk, size, dev, freesize,
2615 verbose);
2616 }
78e44928
NB
2617 /* This is the first device for the array.
2618 * If it is a container, we read it in and do automagic allocations,
2619 * no other devices should be given.
2620 * Otherwise it must be a member device of a container, and we
2621 * do manual allocation.
2622 * Later we should check for a BVD and make an SVD.
a322f70c 2623 */
a322f70c
DW
2624 fd = open(dev, O_RDONLY|O_EXCL, 0);
2625 if (fd >= 0) {
2626 sra = sysfs_read(fd, 0, GET_VERSION);
2627 close(fd);
2628 if (sra && sra->array.major_version == -1 &&
78e44928
NB
2629 strcmp(sra->text_version, "ddf") == 0) {
2630
2631 /* load super */
2632 /* find space for 'n' devices. */
2633 /* remember the devices */
2634 /* Somehow return the fact that we have enough */
a322f70c
DW
2635 }
2636
2c514b71
NB
2637 if (verbose)
2638 fprintf(stderr,
2639 Name ": ddf: Cannot create this array "
b42f577a 2640 "on device %s - a container is required.\n",
2c514b71 2641 dev);
a322f70c
DW
2642 return 0;
2643 }
2644 if (errno != EBUSY || (fd = open(dev, O_RDONLY, 0)) < 0) {
2c514b71
NB
2645 if (verbose)
2646 fprintf(stderr, Name ": ddf: Cannot open %s: %s\n",
2647 dev, strerror(errno));
a322f70c
DW
2648 return 0;
2649 }
2650 /* Well, it is in use by someone, maybe a 'ddf' container. */
2651 cfd = open_container(fd);
2652 if (cfd < 0) {
2653 close(fd);
2c514b71
NB
2654 if (verbose)
2655 fprintf(stderr, Name ": ddf: Cannot use %s: %s\n",
2656 dev, strerror(EBUSY));
a322f70c
DW
2657 return 0;
2658 }
2659 sra = sysfs_read(cfd, 0, GET_VERSION);
2660 close(fd);
2661 if (sra && sra->array.major_version == -1 &&
2662 strcmp(sra->text_version, "ddf") == 0) {
2663 /* This is a member of a ddf container. Load the container
2664 * and try to create a bvd
2665 */
2666 struct ddf_super *ddf;
e1902a7b 2667 if (load_super_ddf_all(st, cfd, (void **)&ddf, NULL) == 0) {
5f8097be 2668 st->sb = ddf;
2f6079dc 2669 st->container_dev = fd2devnum(cfd);
a322f70c 2670 close(cfd);
78e44928 2671 return validate_geometry_ddf_bvd(st, level, layout,
a322f70c 2672 raiddisks, chunk, size,
2c514b71
NB
2673 dev, freesize,
2674 verbose);
a322f70c
DW
2675 }
2676 close(cfd);
c42ec1ed
DW
2677 } else /* device may belong to a different container */
2678 return 0;
2679
a322f70c
DW
2680 return 1;
2681}
2682
2c514b71
NB
2683static int
2684validate_geometry_ddf_container(struct supertype *st,
2685 int level, int layout, int raiddisks,
2686 int chunk, unsigned long long size,
2687 char *dev, unsigned long long *freesize,
2688 int verbose)
a322f70c
DW
2689{
2690 int fd;
2691 unsigned long long ldsize;
2692
2693 if (level != LEVEL_CONTAINER)
2694 return 0;
2695 if (!dev)
2696 return 1;
2697
2698 fd = open(dev, O_RDONLY|O_EXCL, 0);
2699 if (fd < 0) {
2c514b71
NB
2700 if (verbose)
2701 fprintf(stderr, Name ": ddf: Cannot open %s: %s\n",
2702 dev, strerror(errno));
a322f70c
DW
2703 return 0;
2704 }
2705 if (!get_dev_size(fd, dev, &ldsize)) {
2706 close(fd);
2707 return 0;
2708 }
2709 close(fd);
2710
56aca704 2711 *freesize = avail_size_ddf(st, ldsize >> 9);
ea17e7aa
N
2712 if (*freesize == 0)
2713 return 0;
a322f70c
DW
2714
2715 return 1;
2716}
2717
78e44928
NB
2718static int validate_geometry_ddf_bvd(struct supertype *st,
2719 int level, int layout, int raiddisks,
c21e737b 2720 int *chunk, unsigned long long size,
2c514b71
NB
2721 char *dev, unsigned long long *freesize,
2722 int verbose)
a322f70c
DW
2723{
2724 struct stat stb;
2725 struct ddf_super *ddf = st->sb;
2726 struct dl *dl;
5f8097be
NB
2727 unsigned long long pos = 0;
2728 unsigned long long maxsize;
2729 struct extent *e;
2730 int i;
a322f70c 2731 /* ddf/bvd supports lots of things, but not containers */
b42f577a
N
2732 if (level == LEVEL_CONTAINER) {
2733 if (verbose)
2734 fprintf(stderr, Name ": DDF cannot create a container within an container\n");
a322f70c 2735 return 0;
b42f577a 2736 }
a322f70c
DW
2737 /* We must have the container info already read in. */
2738 if (!ddf)
2739 return 0;
2740
5f8097be
NB
2741 if (!dev) {
2742 /* General test: make sure there is space for
2743 * 'raiddisks' device extents of size 'size'.
2744 */
2745 unsigned long long minsize = size;
2746 int dcnt = 0;
2747 if (minsize == 0)
2748 minsize = 8;
2749 for (dl = ddf->dlist; dl ; dl = dl->next)
2750 {
2751 int found = 0;
7e1432fb 2752 pos = 0;
5f8097be
NB
2753
2754 i = 0;
2755 e = get_extents(ddf, dl);
2756 if (!e) continue;
2757 do {
2758 unsigned long long esize;
2759 esize = e[i].start - pos;
2760 if (esize >= minsize)
2761 found = 1;
2762 pos = e[i].start + e[i].size;
2763 i++;
2764 } while (e[i-1].size);
2765 if (found)
2766 dcnt++;
2767 free(e);
2768 }
2769 if (dcnt < raiddisks) {
2c514b71
NB
2770 if (verbose)
2771 fprintf(stderr,
2772 Name ": ddf: Not enough devices with "
2773 "space for this array (%d < %d)\n",
2774 dcnt, raiddisks);
5f8097be
NB
2775 return 0;
2776 }
2777 return 1;
2778 }
a322f70c
DW
2779 /* This device must be a member of the set */
2780 if (stat(dev, &stb) < 0)
2781 return 0;
2782 if ((S_IFMT & stb.st_mode) != S_IFBLK)
2783 return 0;
2784 for (dl = ddf->dlist ; dl ; dl = dl->next) {
f21e18ca
N
2785 if (dl->major == (int)major(stb.st_rdev) &&
2786 dl->minor == (int)minor(stb.st_rdev))
a322f70c
DW
2787 break;
2788 }
5f8097be 2789 if (!dl) {
2c514b71
NB
2790 if (verbose)
2791 fprintf(stderr, Name ": ddf: %s is not in the "
2792 "same DDF set\n",
2793 dev);
5f8097be
NB
2794 return 0;
2795 }
2796 e = get_extents(ddf, dl);
2797 maxsize = 0;
2798 i = 0;
2799 if (e) do {
2800 unsigned long long esize;
2801 esize = e[i].start - pos;
2802 if (esize >= maxsize)
2803 maxsize = esize;
2804 pos = e[i].start + e[i].size;
2805 i++;
2806 } while (e[i-1].size);
2807 *freesize = maxsize;
a322f70c
DW
2808 // FIXME here I am
2809
2810 return 1;
2811}
59e36268 2812
a322f70c 2813static int load_super_ddf_all(struct supertype *st, int fd,
e1902a7b 2814 void **sbp, char *devname)
a322f70c
DW
2815{
2816 struct mdinfo *sra;
2817 struct ddf_super *super;
2818 struct mdinfo *sd, *best = NULL;
2819 int bestseq = 0;
2820 int seq;
2821 char nm[20];
2822 int dfd;
2823
b526e52d 2824 sra = sysfs_read(fd, 0, GET_LEVEL|GET_VERSION|GET_DEVS|GET_STATE);
a322f70c
DW
2825 if (!sra)
2826 return 1;
2827 if (sra->array.major_version != -1 ||
2828 sra->array.minor_version != -2 ||
2829 strcmp(sra->text_version, "ddf") != 0)
2830 return 1;
2831
6416d527 2832 if (posix_memalign((void**)&super, 512, sizeof(*super)) != 0)
a322f70c 2833 return 1;
a2349791 2834 memset(super, 0, sizeof(*super));
a322f70c
DW
2835
2836 /* first, try each device, and choose the best ddf */
2837 for (sd = sra->devs ; sd ; sd = sd->next) {
2838 int rv;
2839 sprintf(nm, "%d:%d", sd->disk.major, sd->disk.minor);
7a7cc504
NB
2840 dfd = dev_open(nm, O_RDONLY);
2841 if (dfd < 0)
a322f70c
DW
2842 return 2;
2843 rv = load_ddf_headers(dfd, super, NULL);
7a7cc504 2844 close(dfd);
a322f70c
DW
2845 if (rv == 0) {
2846 seq = __be32_to_cpu(super->active->seq);
2847 if (super->active->openflag)
2848 seq--;
2849 if (!best || seq > bestseq) {
2850 bestseq = seq;
2851 best = sd;
2852 }
2853 }
2854 }
2855 if (!best)
2856 return 1;
2857 /* OK, load this ddf */
2858 sprintf(nm, "%d:%d", best->disk.major, best->disk.minor);
2859 dfd = dev_open(nm, O_RDONLY);
7a7cc504 2860 if (dfd < 0)
a322f70c
DW
2861 return 1;
2862 load_ddf_headers(dfd, super, NULL);
2863 load_ddf_global(dfd, super, NULL);
2864 close(dfd);
2865 /* Now we need the device-local bits */
2866 for (sd = sra->devs ; sd ; sd = sd->next) {
3d2c4fc7
DW
2867 int rv;
2868
a322f70c 2869 sprintf(nm, "%d:%d", sd->disk.major, sd->disk.minor);
e1902a7b 2870 dfd = dev_open(nm, O_RDWR);
7a7cc504 2871 if (dfd < 0)
a322f70c 2872 return 2;
3d2c4fc7
DW
2873 rv = load_ddf_headers(dfd, super, NULL);
2874 if (rv == 0)
e1902a7b 2875 rv = load_ddf_local(dfd, super, NULL, 1);
3d2c4fc7
DW
2876 if (rv)
2877 return 1;
a322f70c 2878 }
33414a01 2879
a322f70c
DW
2880 *sbp = super;
2881 if (st->ss == NULL) {
78e44928 2882 st->ss = &super_ddf;
a322f70c
DW
2883 st->minor_version = 0;
2884 st->max_devs = 512;
2885 }
47c74f3f 2886 st->container_dev = fd2devnum(fd);
a322f70c
DW
2887 return 0;
2888}
2b959fbf
N
2889
2890static int load_container_ddf(struct supertype *st, int fd,
2891 char *devname)
2892{
2893 return load_super_ddf_all(st, fd, &st->sb, devname);
2894}
2895
0e600426 2896#endif /* MDASSEMBLE */
a322f70c 2897
00bbdbda 2898static struct mdinfo *container_content_ddf(struct supertype *st, char *subarray)
598f0d58
NB
2899{
2900 /* Given a container loaded by load_super_ddf_all,
2901 * extract information about all the arrays into
2902 * an mdinfo tree.
2903 *
2904 * For each vcl in conflist: create an mdinfo, fill it in,
2905 * then look for matching devices (phys_refnum) in dlist
2906 * and create appropriate device mdinfo.
2907 */
2908 struct ddf_super *ddf = st->sb;
2909 struct mdinfo *rest = NULL;
2910 struct vcl *vc;
2911
2912 for (vc = ddf->conflist ; vc ; vc=vc->next)
2913 {
f21e18ca
N
2914 unsigned int i;
2915 unsigned int j;
598f0d58 2916 struct mdinfo *this;
00bbdbda
N
2917 char *ep;
2918
2919 if (subarray &&
2920 (strtoul(subarray, &ep, 10) != vc->vcnum ||
2921 *ep != '\0'))
2922 continue;
2923
598f0d58
NB
2924 this = malloc(sizeof(*this));
2925 memset(this, 0, sizeof(*this));
2926 this->next = rest;
2927 rest = this;
2928
598f0d58
NB
2929 this->array.level = map_num1(ddf_level_num, vc->conf.prl);
2930 this->array.raid_disks =
2931 __be16_to_cpu(vc->conf.prim_elmnt_count);
60f18132
NB
2932 this->array.layout = rlq_to_layout(vc->conf.rlq, vc->conf.prl,
2933 this->array.raid_disks);
598f0d58 2934 this->array.md_minor = -1;
f35f2525
N
2935 this->array.major_version = -1;
2936 this->array.minor_version = -2;
598f0d58
NB
2937 this->array.ctime = DECADE +
2938 __be32_to_cpu(*(__u32*)(vc->conf.guid+16));
2939 this->array.utime = DECADE +
2940 __be32_to_cpu(vc->conf.timestamp);
2941 this->array.chunk_size = 512 << vc->conf.chunk_shift;
2942
59e36268 2943 i = vc->vcnum;
7a7cc504
NB
2944 if ((ddf->virt->entries[i].state & DDF_state_inconsistent) ||
2945 (ddf->virt->entries[i].init_state & DDF_initstate_mask) !=
ed9d66aa 2946 DDF_init_full) {
598f0d58 2947 this->array.state = 0;
ed9d66aa
NB
2948 this->resync_start = 0;
2949 } else {
598f0d58 2950 this->array.state = 1;
b7528a20 2951 this->resync_start = MaxSector;
ed9d66aa 2952 }
db42fa9b
N
2953 memcpy(this->name, ddf->virt->entries[i].name, 16);
2954 this->name[16]=0;
2955 for(j=0; j<16; j++)
2956 if (this->name[j] == ' ')
2957 this->name[j] = 0;
598f0d58
NB
2958
2959 memset(this->uuid, 0, sizeof(this->uuid));
2960 this->component_size = __be64_to_cpu(vc->conf.blocks);
2961 this->array.size = this->component_size / 2;
5f2aace8 2962 this->container_member = i;
598f0d58 2963
c5afc314
N
2964 ddf->currentconf = vc;
2965 uuid_from_super_ddf(st, this->uuid);
2966 ddf->currentconf = NULL;
2967
60f18132
NB
2968 sprintf(this->text_version, "/%s/%d",
2969 devnum2devname(st->container_dev),
2970 this->container_member);
2971
f21e18ca 2972 for (i = 0 ; i < ddf->mppe ; i++) {
598f0d58
NB
2973 struct mdinfo *dev;
2974 struct dl *d;
2975
2976 if (vc->conf.phys_refnum[i] == 0xFFFFFFFF)
2977 continue;
2978
c0d50ca5
N
2979 this->array.working_disks++;
2980
598f0d58
NB
2981 for (d = ddf->dlist; d ; d=d->next)
2982 if (d->disk.refnum == vc->conf.phys_refnum[i])
2983 break;
2984 if (d == NULL)
bc17324f
N
2985 /* Haven't found that one yet, maybe there are others */
2986 continue;
2987
598f0d58
NB
2988 dev = malloc(sizeof(*dev));
2989 memset(dev, 0, sizeof(*dev));
2990 dev->next = this->devs;
2991 this->devs = dev;
2992
2993 dev->disk.number = __be32_to_cpu(d->disk.refnum);
2994 dev->disk.major = d->major;
2995 dev->disk.minor = d->minor;
2996 dev->disk.raid_disk = i;
2997 dev->disk.state = (1<<MD_DISK_SYNC)|(1<<MD_DISK_ACTIVE);
d23534e4 2998 dev->recovery_start = MaxSector;
598f0d58 2999
120f7677
NB
3000 dev->events = __be32_to_cpu(ddf->primary.seq);
3001 dev->data_offset = __be64_to_cpu(vc->lba_offset[i]);
598f0d58
NB
3002 dev->component_size = __be64_to_cpu(vc->conf.blocks);
3003 if (d->devname)
3004 strcpy(dev->name, d->devname);
3005 }
3006 }
3007 return rest;
3008}
3009
955e9ea1 3010static int store_super_ddf(struct supertype *st, int fd)
a322f70c 3011{
955e9ea1 3012 struct ddf_super *ddf = st->sb;
a322f70c 3013 unsigned long long dsize;
6416d527 3014 void *buf;
3d2c4fc7 3015 int rc;
a322f70c 3016
955e9ea1
DW
3017 if (!ddf)
3018 return 1;
3019
3020 /* ->dlist and ->conflist will be set for updates, currently not
3021 * supported
3022 */
3023 if (ddf->dlist || ddf->conflist)
3024 return 1;
3025
a322f70c
DW
3026 if (!get_dev_size(fd, NULL, &dsize))
3027 return 1;
3028
3d2c4fc7
DW
3029 if (posix_memalign(&buf, 512, 512) != 0)
3030 return 1;
6416d527
NB
3031 memset(buf, 0, 512);
3032
a322f70c 3033 lseek64(fd, dsize-512, 0);
3d2c4fc7 3034 rc = write(fd, buf, 512);
6416d527 3035 free(buf);
3d2c4fc7
DW
3036 if (rc < 0)
3037 return 1;
a322f70c
DW
3038 return 0;
3039}
3040
a19c88b8
NB
3041static int compare_super_ddf(struct supertype *st, struct supertype *tst)
3042{
3043 /*
3044 * return:
3045 * 0 same, or first was empty, and second was copied
3046 * 1 second had wrong number
3047 * 2 wrong uuid
3048 * 3 wrong other info
3049 */
3050 struct ddf_super *first = st->sb;
3051 struct ddf_super *second = tst->sb;
3052
3053 if (!first) {
3054 st->sb = tst->sb;
3055 tst->sb = NULL;
3056 return 0;
3057 }
3058
3059 if (memcmp(first->anchor.guid, second->anchor.guid, DDF_GUID_LEN) != 0)
3060 return 2;
3061
3062 /* FIXME should I look at anything else? */
3063 return 0;
3064}
3065
0e600426 3066#ifndef MDASSEMBLE
4e5528c6
NB
3067/*
3068 * A new array 'a' has been started which claims to be instance 'inst'
3069 * within container 'c'.
3070 * We need to confirm that the array matches the metadata in 'c' so
3071 * that we don't corrupt any metadata.
3072 */
cba0191b 3073static int ddf_open_new(struct supertype *c, struct active_array *a, char *inst)
549e9569 3074{
2c514b71 3075 dprintf("ddf: open_new %s\n", inst);
cba0191b 3076 a->info.container_member = atoi(inst);
549e9569
NB
3077 return 0;
3078}
3079
4e5528c6
NB
3080/*
3081 * The array 'a' is to be marked clean in the metadata.
ed9d66aa 3082 * If '->resync_start' is not ~(unsigned long long)0, then the array is only
4e5528c6
NB
3083 * clean up to the point (in sectors). If that cannot be recorded in the
3084 * metadata, then leave it as dirty.
3085 *
3086 * For DDF, we need to clear the DDF_state_inconsistent bit in the
3087 * !global! virtual_disk.virtual_entry structure.
3088 */
01f157d7 3089static int ddf_set_array_state(struct active_array *a, int consistent)
549e9569 3090{
4e5528c6
NB
3091 struct ddf_super *ddf = a->container->sb;
3092 int inst = a->info.container_member;
18a2f463 3093 int old = ddf->virt->entries[inst].state;
01f157d7
N
3094 if (consistent == 2) {
3095 /* Should check if a recovery should be started FIXME */
3096 consistent = 1;
b7941fd6 3097 if (!is_resync_complete(&a->info))
01f157d7
N
3098 consistent = 0;
3099 }
ed9d66aa
NB
3100 if (consistent)
3101 ddf->virt->entries[inst].state &= ~DDF_state_inconsistent;
3102 else
4e5528c6 3103 ddf->virt->entries[inst].state |= DDF_state_inconsistent;
18a2f463
NB
3104 if (old != ddf->virt->entries[inst].state)
3105 ddf->updates_pending = 1;
3106
3107 old = ddf->virt->entries[inst].init_state;
ed9d66aa 3108 ddf->virt->entries[inst].init_state &= ~DDF_initstate_mask;
b7941fd6 3109 if (is_resync_complete(&a->info))
ed9d66aa 3110 ddf->virt->entries[inst].init_state |= DDF_init_full;
b7941fd6 3111 else if (a->info.resync_start == 0)
ed9d66aa 3112 ddf->virt->entries[inst].init_state |= DDF_init_not;
4e5528c6 3113 else
ed9d66aa 3114 ddf->virt->entries[inst].init_state |= DDF_init_quick;
18a2f463
NB
3115 if (old != ddf->virt->entries[inst].init_state)
3116 ddf->updates_pending = 1;
ed9d66aa 3117
2c514b71 3118 dprintf("ddf mark %d %s %llu\n", inst, consistent?"clean":"dirty",
b7941fd6 3119 a->info.resync_start);
01f157d7 3120 return consistent;
fd7cde1b
DW
3121}
3122
7a7cc504
NB
3123/*
3124 * The state of each disk is stored in the global phys_disk structure
3125 * in phys_disk.entries[n].state.
3126 * This makes various combinations awkward.
3127 * - When a device fails in any array, it must be failed in all arrays
3128 * that include a part of this device.
3129 * - When a component is rebuilding, we cannot include it officially in the
3130 * array unless this is the only array that uses the device.
3131 *
3132 * So: when transitioning:
3133 * Online -> failed, just set failed flag. monitor will propagate
3134 * spare -> online, the device might need to be added to the array.
3135 * spare -> failed, just set failed. Don't worry if in array or not.
3136 */
8d45d196 3137static void ddf_set_disk(struct active_array *a, int n, int state)
549e9569 3138{
7a7cc504 3139 struct ddf_super *ddf = a->container->sb;
f21e18ca 3140 unsigned int inst = a->info.container_member;
7a7cc504
NB
3141 struct vd_config *vc = find_vdcr(ddf, inst);
3142 int pd = find_phys(ddf, vc->phys_refnum[n]);
3143 int i, st, working;
3144
3145 if (vc == NULL) {
2c514b71 3146 dprintf("ddf: cannot find instance %d!!\n", inst);
7a7cc504
NB
3147 return;
3148 }
3149 if (pd < 0) {
3150 /* disk doesn't currently exist. If it is now in_sync,
3151 * insert it. */
3152 if ((state & DS_INSYNC) && ! (state & DS_FAULTY)) {
3153 /* Find dev 'n' in a->info->devs, determine the
3154 * ddf refnum, and set vc->phys_refnum and update
3155 * phys->entries[]
3156 */
3157 /* FIXME */
3158 }
3159 } else {
18a2f463 3160 int old = ddf->phys->entries[pd].state;
7a7cc504
NB
3161 if (state & DS_FAULTY)
3162 ddf->phys->entries[pd].state |= __cpu_to_be16(DDF_Failed);
3163 if (state & DS_INSYNC) {
3164 ddf->phys->entries[pd].state |= __cpu_to_be16(DDF_Online);
3165 ddf->phys->entries[pd].state &= __cpu_to_be16(~DDF_Rebuilding);
3166 }
18a2f463
NB
3167 if (old != ddf->phys->entries[pd].state)
3168 ddf->updates_pending = 1;
7a7cc504
NB
3169 }
3170
2c514b71 3171 dprintf("ddf: set_disk %d to %x\n", n, state);
7e1432fb 3172
7a7cc504
NB
3173 /* Now we need to check the state of the array and update
3174 * virtual_disk.entries[n].state.
3175 * It needs to be one of "optimal", "degraded", "failed".
3176 * I don't understand 'deleted' or 'missing'.
3177 */
3178 working = 0;
3179 for (i=0; i < a->info.array.raid_disks; i++) {
3180 pd = find_phys(ddf, vc->phys_refnum[i]);
3181 if (pd < 0)
3182 continue;
57632f4a
NB
3183 st = __be16_to_cpu(ddf->phys->entries[pd].state);
3184 if ((st & (DDF_Online|DDF_Failed|DDF_Rebuilding))
7a7cc504
NB
3185 == DDF_Online)
3186 working++;
3187 }
3188 state = DDF_state_degraded;
3189 if (working == a->info.array.raid_disks)
3190 state = DDF_state_optimal;
3191 else switch(vc->prl) {
3192 case DDF_RAID0:
3193 case DDF_CONCAT:
3194 case DDF_JBOD:
3195 state = DDF_state_failed;
3196 break;
3197 case DDF_RAID1:
3198 if (working == 0)
3199 state = DDF_state_failed;
3200 break;
3201 case DDF_RAID4:
3202 case DDF_RAID5:
3203 if (working < a->info.array.raid_disks-1)
3204 state = DDF_state_failed;
3205 break;
3206 case DDF_RAID6:
3207 if (working < a->info.array.raid_disks-2)
3208 state = DDF_state_failed;
3209 else if (working == a->info.array.raid_disks-1)
3210 state = DDF_state_part_optimal;
3211 break;
3212 }
3213
18a2f463
NB
3214 if (ddf->virt->entries[inst].state !=
3215 ((ddf->virt->entries[inst].state & ~DDF_state_mask)
3216 | state)) {
3217
3218 ddf->virt->entries[inst].state =
3219 (ddf->virt->entries[inst].state & ~DDF_state_mask)
3220 | state;
3221 ddf->updates_pending = 1;
3222 }
7a7cc504 3223
549e9569
NB
3224}
3225
2e735d19 3226static void ddf_sync_metadata(struct supertype *st)
549e9569 3227{
7a7cc504
NB
3228
3229 /*
3230 * Write all data to all devices.
3231 * Later, we might be able to track whether only local changes
3232 * have been made, or whether any global data has been changed,
3233 * but ddf is sufficiently weird that it probably always
3234 * changes global data ....
3235 */
18a2f463
NB
3236 struct ddf_super *ddf = st->sb;
3237 if (!ddf->updates_pending)
3238 return;
3239 ddf->updates_pending = 0;
1cc7f4fe 3240 __write_init_super_ddf(st);
2c514b71 3241 dprintf("ddf: sync_metadata\n");
549e9569
NB
3242}
3243
88c164f4
NB
3244static void ddf_process_update(struct supertype *st,
3245 struct metadata_update *update)
3246{
3247 /* Apply this update to the metadata.
3248 * The first 4 bytes are a DDF_*_MAGIC which guides
3249 * our actions.
3250 * Possible update are:
3251 * DDF_PHYS_RECORDS_MAGIC
3252 * Add a new physical device. Changes to this record
3253 * only happen implicitly.
3254 * used_pdes is the device number.
3255 * DDF_VIRT_RECORDS_MAGIC
3256 * Add a new VD. Possibly also change the 'access' bits.
3257 * populated_vdes is the entry number.
3258 * DDF_VD_CONF_MAGIC
3259 * New or updated VD. the VIRT_RECORD must already
3260 * exist. For an update, phys_refnum and lba_offset
3261 * (at least) are updated, and the VD_CONF must
3262 * be written to precisely those devices listed with
3263 * a phys_refnum.
3264 * DDF_SPARE_ASSIGN_MAGIC
3265 * replacement Spare Assignment Record... but for which device?
3266 *
3267 * So, e.g.:
3268 * - to create a new array, we send a VIRT_RECORD and
3269 * a VD_CONF. Then assemble and start the array.
3270 * - to activate a spare we send a VD_CONF to add the phys_refnum
3271 * and offset. This will also mark the spare as active with
3272 * a spare-assignment record.
3273 */
3274 struct ddf_super *ddf = st->sb;
3275 __u32 *magic = (__u32*)update->buf;
3276 struct phys_disk *pd;
3277 struct virtual_disk *vd;
3278 struct vd_config *vc;
3279 struct vcl *vcl;
3280 struct dl *dl;
f21e18ca
N
3281 unsigned int mppe;
3282 unsigned int ent;
88c164f4 3283
2c514b71 3284 dprintf("Process update %x\n", *magic);
7e1432fb 3285
88c164f4
NB
3286 switch (*magic) {
3287 case DDF_PHYS_RECORDS_MAGIC:
3288
3289 if (update->len != (sizeof(struct phys_disk) +
3290 sizeof(struct phys_disk_entry)))
3291 return;
3292 pd = (struct phys_disk*)update->buf;
3293
3294 ent = __be16_to_cpu(pd->used_pdes);
3295 if (ent >= __be16_to_cpu(ddf->phys->max_pdes))
3296 return;
3297 if (!all_ff(ddf->phys->entries[ent].guid))
3298 return;
3299 ddf->phys->entries[ent] = pd->entries[0];
3300 ddf->phys->used_pdes = __cpu_to_be16(1 +
3301 __be16_to_cpu(ddf->phys->used_pdes));
18a2f463 3302 ddf->updates_pending = 1;
2cc2983d
N
3303 if (ddf->add_list) {
3304 struct active_array *a;
3305 struct dl *al = ddf->add_list;
3306 ddf->add_list = al->next;
3307
3308 al->next = ddf->dlist;
3309 ddf->dlist = al;
3310
3311 /* As a device has been added, we should check
3312 * for any degraded devices that might make
3313 * use of this spare */
3314 for (a = st->arrays ; a; a=a->next)
3315 a->check_degraded = 1;
3316 }
88c164f4
NB
3317 break;
3318
3319 case DDF_VIRT_RECORDS_MAGIC:
3320
3321 if (update->len != (sizeof(struct virtual_disk) +
3322 sizeof(struct virtual_entry)))
3323 return;
3324 vd = (struct virtual_disk*)update->buf;
3325
3326 ent = __be16_to_cpu(vd->populated_vdes);
3327 if (ent >= __be16_to_cpu(ddf->virt->max_vdes))
3328 return;
3329 if (!all_ff(ddf->virt->entries[ent].guid))
3330 return;
3331 ddf->virt->entries[ent] = vd->entries[0];
3332 ddf->virt->populated_vdes = __cpu_to_be16(1 +
3333 __be16_to_cpu(ddf->virt->populated_vdes));
18a2f463 3334 ddf->updates_pending = 1;
88c164f4
NB
3335 break;
3336
3337 case DDF_VD_CONF_MAGIC:
2c514b71 3338 dprintf("len %d %d\n", update->len, ddf->conf_rec_len);
88c164f4
NB
3339
3340 mppe = __be16_to_cpu(ddf->anchor.max_primary_element_entries);
f21e18ca 3341 if ((unsigned)update->len != ddf->conf_rec_len * 512)
88c164f4
NB
3342 return;
3343 vc = (struct vd_config*)update->buf;
3344 for (vcl = ddf->conflist; vcl ; vcl = vcl->next)
3345 if (memcmp(vcl->conf.guid, vc->guid, DDF_GUID_LEN) == 0)
3346 break;
2c514b71 3347 dprintf("vcl = %p\n", vcl);
88c164f4
NB
3348 if (vcl) {
3349 /* An update, just copy the phys_refnum and lba_offset
3350 * fields
3351 */
3352 memcpy(vcl->conf.phys_refnum, vc->phys_refnum,
3353 mppe * (sizeof(__u32) + sizeof(__u64)));
3354 } else {
3355 /* A new VD_CONF */
e6b9548d
DW
3356 if (!update->space)
3357 return;
88c164f4
NB
3358 vcl = update->space;
3359 update->space = NULL;
3360 vcl->next = ddf->conflist;
edd8d13c 3361 memcpy(&vcl->conf, vc, update->len);
88c164f4
NB
3362 vcl->lba_offset = (__u64*)
3363 &vcl->conf.phys_refnum[mppe];
3364 ddf->conflist = vcl;
3365 }
3366 /* Now make sure vlist is correct for each dl. */
3367 for (dl = ddf->dlist; dl; dl = dl->next) {
f21e18ca
N
3368 unsigned int dn;
3369 unsigned int vn = 0;
88c164f4
NB
3370 for (vcl = ddf->conflist; vcl ; vcl = vcl->next)
3371 for (dn=0; dn < ddf->mppe ; dn++)
3372 if (vcl->conf.phys_refnum[dn] ==
3373 dl->disk.refnum) {
2c514b71
NB
3374 dprintf("dev %d has %p at %d\n",
3375 dl->pdnum, vcl, vn);
88c164f4
NB
3376 dl->vlist[vn++] = vcl;
3377 break;
3378 }
3379 while (vn < ddf->max_part)
3380 dl->vlist[vn++] = NULL;
7e1432fb
NB
3381 if (dl->vlist[0]) {
3382 ddf->phys->entries[dl->pdnum].type &=
3383 ~__cpu_to_be16(DDF_Global_Spare);
3384 ddf->phys->entries[dl->pdnum].type |=
3385 __cpu_to_be16(DDF_Active_in_VD);
3386 }
3387 if (dl->spare) {
3388 ddf->phys->entries[dl->pdnum].type &=
3389 ~__cpu_to_be16(DDF_Global_Spare);
3390 ddf->phys->entries[dl->pdnum].type |=
3391 __cpu_to_be16(DDF_Spare);
3392 }
3393 if (!dl->vlist[0] && !dl->spare) {
3394 ddf->phys->entries[dl->pdnum].type |=
3395 __cpu_to_be16(DDF_Global_Spare);
3396 ddf->phys->entries[dl->pdnum].type &=
3397 ~__cpu_to_be16(DDF_Spare |
3398 DDF_Active_in_VD);
3399 }
88c164f4 3400 }
18a2f463 3401 ddf->updates_pending = 1;
88c164f4
NB
3402 break;
3403 case DDF_SPARE_ASSIGN_MAGIC:
3404 default: break;
3405 }
3406}
3407
edd8d13c
NB
3408static void ddf_prepare_update(struct supertype *st,
3409 struct metadata_update *update)
3410{
3411 /* This update arrived at managemon.
3412 * We are about to pass it to monitor.
3413 * If a malloc is needed, do it here.
3414 */
3415 struct ddf_super *ddf = st->sb;
3416 __u32 *magic = (__u32*)update->buf;
3417 if (*magic == DDF_VD_CONF_MAGIC)
e6b9548d 3418 if (posix_memalign(&update->space, 512,
6416d527 3419 offsetof(struct vcl, conf)
e6b9548d
DW
3420 + ddf->conf_rec_len * 512) != 0)
3421 update->space = NULL;
edd8d13c
NB
3422}
3423
7e1432fb
NB
3424/*
3425 * Check if the array 'a' is degraded but not failed.
3426 * If it is, find as many spares as are available and needed and
3427 * arrange for their inclusion.
3428 * We only choose devices which are not already in the array,
3429 * and prefer those with a spare-assignment to this array.
3430 * otherwise we choose global spares - assuming always that
3431 * there is enough room.
3432 * For each spare that we assign, we return an 'mdinfo' which
3433 * describes the position for the device in the array.
3434 * We also add to 'updates' a DDF_VD_CONF_MAGIC update with
3435 * the new phys_refnum and lba_offset values.
3436 *
3437 * Only worry about BVDs at the moment.
3438 */
3439static struct mdinfo *ddf_activate_spare(struct active_array *a,
3440 struct metadata_update **updates)
3441{
3442 int working = 0;
3443 struct mdinfo *d;
3444 struct ddf_super *ddf = a->container->sb;
3445 int global_ok = 0;
3446 struct mdinfo *rv = NULL;
3447 struct mdinfo *di;
3448 struct metadata_update *mu;
3449 struct dl *dl;
3450 int i;
3451 struct vd_config *vc;
3452 __u64 *lba;
3453
7e1432fb
NB
3454 for (d = a->info.devs ; d ; d = d->next) {
3455 if ((d->curr_state & DS_FAULTY) &&
3456 d->state_fd >= 0)
3457 /* wait for Removal to happen */
3458 return NULL;
3459 if (d->state_fd >= 0)
3460 working ++;
3461 }
3462
2c514b71
NB
3463 dprintf("ddf_activate: working=%d (%d) level=%d\n", working, a->info.array.raid_disks,
3464 a->info.array.level);
7e1432fb
NB
3465 if (working == a->info.array.raid_disks)
3466 return NULL; /* array not degraded */
3467 switch (a->info.array.level) {
3468 case 1:
3469 if (working == 0)
3470 return NULL; /* failed */
3471 break;
3472 case 4:
3473 case 5:
3474 if (working < a->info.array.raid_disks - 1)
3475 return NULL; /* failed */
3476 break;
3477 case 6:
3478 if (working < a->info.array.raid_disks - 2)
3479 return NULL; /* failed */
3480 break;
3481 default: /* concat or stripe */
3482 return NULL; /* failed */
3483 }
3484
3485 /* For each slot, if it is not working, find a spare */
3486 dl = ddf->dlist;
3487 for (i = 0; i < a->info.array.raid_disks; i++) {
3488 for (d = a->info.devs ; d ; d = d->next)
3489 if (d->disk.raid_disk == i)
3490 break;
2c514b71 3491 dprintf("found %d: %p %x\n", i, d, d?d->curr_state:0);
7e1432fb
NB
3492 if (d && (d->state_fd >= 0))
3493 continue;
3494
3495 /* OK, this device needs recovery. Find a spare */
3496 again:
3497 for ( ; dl ; dl = dl->next) {
3498 unsigned long long esize;
3499 unsigned long long pos;
3500 struct mdinfo *d2;
3501 int is_global = 0;
3502 int is_dedicated = 0;
3503 struct extent *ex;
f21e18ca 3504 unsigned int j;
7e1432fb
NB
3505 /* If in this array, skip */
3506 for (d2 = a->info.devs ; d2 ; d2 = d2->next)
3507 if (d2->disk.major == dl->major &&
3508 d2->disk.minor == dl->minor) {
2c514b71 3509 dprintf("%x:%x already in array\n", dl->major, dl->minor);
7e1432fb
NB
3510 break;
3511 }
3512 if (d2)
3513 continue;
3514 if (ddf->phys->entries[dl->pdnum].type &
3515 __cpu_to_be16(DDF_Spare)) {
3516 /* Check spare assign record */
3517 if (dl->spare) {
3518 if (dl->spare->type & DDF_spare_dedicated) {
3519 /* check spare_ents for guid */
3520 for (j = 0 ;
3521 j < __be16_to_cpu(dl->spare->populated);
3522 j++) {
3523 if (memcmp(dl->spare->spare_ents[j].guid,
3524 ddf->virt->entries[a->info.container_member].guid,
3525 DDF_GUID_LEN) == 0)
3526 is_dedicated = 1;
3527 }
3528 } else
3529 is_global = 1;
3530 }
3531 } else if (ddf->phys->entries[dl->pdnum].type &
3532 __cpu_to_be16(DDF_Global_Spare)) {
3533 is_global = 1;
3534 }
3535 if ( ! (is_dedicated ||
3536 (is_global && global_ok))) {
2c514b71 3537 dprintf("%x:%x not suitable: %d %d\n", dl->major, dl->minor,
7e1432fb
NB
3538 is_dedicated, is_global);
3539 continue;
3540 }
3541
3542 /* We are allowed to use this device - is there space?
3543 * We need a->info.component_size sectors */
3544 ex = get_extents(ddf, dl);
3545 if (!ex) {
2c514b71 3546 dprintf("cannot get extents\n");
7e1432fb
NB
3547 continue;
3548 }
3549 j = 0; pos = 0;
3550 esize = 0;
3551
3552 do {
3553 esize = ex[j].start - pos;
3554 if (esize >= a->info.component_size)
3555 break;
3556 pos = ex[i].start + ex[i].size;
3557 i++;
3558 } while (ex[i-1].size);
3559
3560 free(ex);
3561 if (esize < a->info.component_size) {
2c514b71
NB
3562 dprintf("%x:%x has no room: %llu %llu\n", dl->major, dl->minor,
3563 esize, a->info.component_size);
7e1432fb
NB
3564 /* No room */
3565 continue;
3566 }
3567
3568 /* Cool, we have a device with some space at pos */
3569 di = malloc(sizeof(*di));
79244939
DW
3570 if (!di)
3571 continue;
7e1432fb
NB
3572 memset(di, 0, sizeof(*di));
3573 di->disk.number = i;
3574 di->disk.raid_disk = i;
3575 di->disk.major = dl->major;
3576 di->disk.minor = dl->minor;
3577 di->disk.state = 0;
d23534e4 3578 di->recovery_start = 0;
7e1432fb
NB
3579 di->data_offset = pos;
3580 di->component_size = a->info.component_size;
3581 di->container_member = dl->pdnum;
3582 di->next = rv;
3583 rv = di;
2c514b71
NB
3584 dprintf("%x:%x to be %d at %llu\n", dl->major, dl->minor,
3585 i, pos);
7e1432fb
NB
3586
3587 break;
3588 }
3589 if (!dl && ! global_ok) {
3590 /* not enough dedicated spares, try global */
3591 global_ok = 1;
3592 dl = ddf->dlist;
3593 goto again;
3594 }
3595 }
3596
3597 if (!rv)
3598 /* No spares found */
3599 return rv;
3600 /* Now 'rv' has a list of devices to return.
3601 * Create a metadata_update record to update the
3602 * phys_refnum and lba_offset values
3603 */
904c1ef7 3604 mu = malloc(sizeof(*mu));
79244939
DW
3605 if (mu && posix_memalign(&mu->space, 512, sizeof(struct vcl)) != 0) {
3606 free(mu);
3607 mu = NULL;
3608 }
3609 if (!mu) {
3610 while (rv) {
3611 struct mdinfo *n = rv->next;
3612
3613 free(rv);
3614 rv = n;
3615 }
3616 return NULL;
3617 }
3618
904c1ef7 3619 mu->buf = malloc(ddf->conf_rec_len * 512);
7e1432fb
NB
3620 mu->len = ddf->conf_rec_len;
3621 mu->next = *updates;
3622 vc = find_vdcr(ddf, a->info.container_member);
3623 memcpy(mu->buf, vc, ddf->conf_rec_len * 512);
3624
3625 vc = (struct vd_config*)mu->buf;
3626 lba = (__u64*)&vc->phys_refnum[ddf->mppe];
3627 for (di = rv ; di ; di = di->next) {
3628 vc->phys_refnum[di->disk.raid_disk] =
3629 ddf->phys->entries[dl->pdnum].refnum;
3630 lba[di->disk.raid_disk] = di->data_offset;
3631 }
3632 *updates = mu;
3633 return rv;
3634}
0e600426 3635#endif /* MDASSEMBLE */
7e1432fb 3636
b640a252
N
3637static int ddf_level_to_layout(int level)
3638{
3639 switch(level) {
3640 case 0:
3641 case 1:
3642 return 0;
3643 case 5:
3644 return ALGORITHM_LEFT_SYMMETRIC;
3645 case 6:
3646 return ALGORITHM_ROTATING_N_CONTINUE;
3647 case 10:
3648 return 0x102;
3649 default:
3650 return UnSet;
3651 }
3652}
3653
30f58b22
DW
3654static void default_geometry_ddf(struct supertype *st, int *level, int *layout, int *chunk)
3655{
3656 if (level && *level == UnSet)
3657 *level = LEVEL_CONTAINER;
3658
3659 if (level && layout && *layout == UnSet)
3660 *layout = ddf_level_to_layout(*level);
3661}
3662
a322f70c
DW
3663struct superswitch super_ddf = {
3664#ifndef MDASSEMBLE
3665 .examine_super = examine_super_ddf,
3666 .brief_examine_super = brief_examine_super_ddf,
4737ae25 3667 .brief_examine_subarrays = brief_examine_subarrays_ddf,
bceedeec 3668 .export_examine_super = export_examine_super_ddf,
a322f70c
DW
3669 .detail_super = detail_super_ddf,
3670 .brief_detail_super = brief_detail_super_ddf,
3671 .validate_geometry = validate_geometry_ddf,
78e44928 3672 .write_init_super = write_init_super_ddf,
0e600426 3673 .add_to_super = add_to_super_ddf,
2b959fbf 3674 .load_container = load_container_ddf,
a322f70c
DW
3675#endif
3676 .match_home = match_home_ddf,
3677 .uuid_from_super= uuid_from_super_ddf,
3678 .getinfo_super = getinfo_super_ddf,
3679 .update_super = update_super_ddf,
3680
3681 .avail_size = avail_size_ddf,
3682
a19c88b8
NB
3683 .compare_super = compare_super_ddf,
3684
a322f70c 3685 .load_super = load_super_ddf,
ba7eb04f 3686 .init_super = init_super_ddf,
955e9ea1 3687 .store_super = store_super_ddf,
a322f70c
DW
3688 .free_super = free_super_ddf,
3689 .match_metadata_desc = match_metadata_desc_ddf,
78e44928 3690 .container_content = container_content_ddf,
30f58b22 3691 .default_geometry = default_geometry_ddf,
a322f70c 3692
a322f70c 3693 .external = 1,
549e9569 3694
0e600426 3695#ifndef MDASSEMBLE
549e9569
NB
3696/* for mdmon */
3697 .open_new = ddf_open_new,
ed9d66aa 3698 .set_array_state= ddf_set_array_state,
549e9569
NB
3699 .set_disk = ddf_set_disk,
3700 .sync_metadata = ddf_sync_metadata,
88c164f4 3701 .process_update = ddf_process_update,
edd8d13c 3702 .prepare_update = ddf_prepare_update,
7e1432fb 3703 .activate_spare = ddf_activate_spare,
0e600426 3704#endif
4cce4069 3705 .name = "ddf",
a322f70c 3706};